Hi, I’m the author of this blog post. I submitted it to HN two weeks ago, but it didn’t gain much traction, so I’m glad to see it resubmitted. Happy to answer any questions, though I think the situation is fairly self-explanatory.
Using datasets from the open web archive Common Crawl,
we identified and analyzed 159,041 Lovable projects from
December 2025 to September 2026. Of these, 34,545 used
Supabase.
24,664 websites sent database records from Supabase as
soon as they loaded, without requiring a login or any
other interaction. That alone is not a security issue:
the data may be intended for public display. But by
making additional read-only requests, we were able to
retrieve more database content. We found records
containing email addresses in 10,560 tables across 7,403
of the 24,664 websites (30%).
What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data. Using a confidence threshold of 85%, we found that nearly 4,000 websites leaked such information.
Edit: That's 15,9% of all Lovable websites that use Supabase.
Don't you think you should have reported this after you established the principle rather than to download more than you needed to do so? Because it looks to me as though you went from 'white hat hacker' to 'quite probably criminal' without a really good reason.
Do you think Lovable would be transparent as to the extent of the problem? Or would they use terminology like “a small subset of our users” and “data not including card details” or whatever to minimize the issue.
I’d much rather live in a world where independent researchers can figure out the extent of the issue rather than go “uh oh there might be an issue here, guess we’ll never know the extent of it, let’s trust them to fix it”.
That's a good question. But it is a very fine line between being a white hat that reports a vulnerability in good faith and someone who actually exploits that vulnerability and turns a vulnerability report into an actual breach. This is - in my opinion, clearly the HN hive mind disagrees - across that line.
For instance, if I run into a server and I see 'passport.jpg' in a directory listing I could download it to see if it really is a passport. But at that point in time I am flirting with becoming a defendant in a data breach rather than that I'm reporting a possible vulnerability. Sure, you could argue that the site owner might close the hole and then they could downplay that it really was a passport photograph. But I care more about the hole being closed than that I care about the contents of the file, which clearly are not meant to be disclosed based on the filename alone.
Independent researchers that are under contract with the companies they review sometimes have more leeway in what they are allowed to do. But downloading enough data to be able to run statistical analysis using LLMs is turning this into a reportable breach and that is a great way to get this whole thing to backfire, possibly to the point where the OP turns defendant.
the article mentions "we retrieved only parts of the datasets rather than downloading them in full", so a charitable interpretation might be that for each identified source candidate, they spot checked (or had a model spot check) a few individual records to have incontrovertible proof of breach of privacy, and then ran an aggregate count query to gauge the extent of breach. I think that would be fairly in line with "white hat" behavior that's doing the smallest amount of extra work towards a solid proof and clearly not taking the obvious and easy next step of downloading the remainder that would be required for self enrichment. Difficult to argue that there is an intention of malice.
A bad actor would be downloading /all/ the data and disclosing it to no one, and selling it while obscuring the source.
That said I agree that that certainly might fall afoul of the letter of whatever local laws apply, and certainly prosecutors can also be over-zealous in their interpretations. So personally I would never risk doing something that comes up against that line, from a personal risk POV. But idk if it's fair to argue that this is NOT good faith.
I don't argue it is not good faith. I just wonder if OP realized that they might be crossing a line they shouldn't be crossing, ever, while doing security research without being under contract to the party they are investigating.
Then at least you are under and NDA and there is no way for the company to spin it that you were a bad guy.
The numbers mentioned are pretty serious: "We found records
containing email addresses in 10,560 tables across 7,403
of the 24,664 websites (30%)."
To me - and please correct me if I'm wrong - that reads like 'we downloaded at least parts of 10560 tables'.
How do you know which websites are affected without sampling data from each website in question? As the blog explains, it's safe to serve public data from Supabase in this manner, so the only way to determine whether the website has an issue is to look at the actual records they are serving. Their approach seems entirely reasonable to me.
As someone who has dealt with vulnerability reports on products before, I'd strongly prefer getting concrete proof of an issue versus "your product came back as positive on our vulnerability scanner^tm" which is 99% of the time going to be a scam or a waste of time.
Yes, would you like the report to be what you need to know or a report that will cause you to have to report an actual breach rather than just a vulnerability? Because that is what this is.
How does that relate to e.g. have I been pwnd? If the obligation to be a white hat is to protect the owners of the information then is it not also to inform the persons who have been impacted a valid white hat activity? From my perspective the classification of someone as a black hat should be if they use it for something malicious. And if anything disclosure limits any potential use. On the other hand our big LLM companies Hoover over everything, have access to it, occasionally go rogue with their knowledge and access, but I don’t see many people calling them black hats.
You'll notice that the person that built have I been pwnd has been extremely careful in how the data comes into their possession, I highly doubt they would ever go so far as to breach a site themselves. It would go completely against their mission to do so.
> On the other hand our big LLM companies Hoover over everything, have access to it, occasionally go rogue with their knowledge and access, but I don’t see many people calling them black hats.
I'll be more than happy to call them criminals and I've warned very publicly about their resetting configuration switches to allow them to pretend that they have your consent to use your data for training purposes.
That's an important bit of information that probably should make it into the blog, and I hope you guys made the right call, but I think that that is probably not the case. Best of luck.
oh: "What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data."
I hope at least that that was a local model. If not you've made things considerably worse.
It’s mentioned in the Handelsblatt article linked in the blog post. Since the lawyer was brought in by the newspaper, we didn’t mention their involvement in the post.
All LLM-based analysis was performed using local models, specifically the "elephant-1" model on our local RTX Pro 4000 SFF.
I respect your insistence on proper data handling -- you’re right in doing so -- and we tried to be as careful as possible when conducting the analysis.
The lawyer that was brought in by the newspaper was making sure the newspaper wouldn't end up being liable. The only lawyer whose advice you should take is the one that you pay.
As long as they are representing someone I think lawyers have the same obligations regardless of who actually pays them or if they are receiving any payment at all.
What I mean is: you knew it worked before you went and pulled more data, you could have stopped after the first two records, why pull more than you needed to prove that it was possible?
We retrieved only enough data to determine whether it was privacy-sensitive, which could not be reliably assessed from just two records. We did not download any more than that.
Supabase tables also use pagination, and we never downloaded more than the first page.
I may not be the intended audience, but as the author of a similar system having a benchmark like this of what user behavior can result in when it’s unconstrained is helpful for my safety evals. So in an indirect way, you’re making future platforms more secure by sharing this.
You probably should not have done that. It's one thing to determine that a hole exists, an entirely different one to exploit it to the point that you can do damage assessment on the target.
You seem to be insanely hellbent on implying the actually worst possible intent to the researcher who acted in the good faith and contacted the vendor aiming for the quickest and best fix.
It looks like you're weirdly trying to inflate a legitimate attempt to confirm if the severity of the exposure is as bad as suspected into a full blown nasty data thief.
This is why everyone is so puzzled. You're genuinely harassing op for doing a good thing and responsibly reporting the severe security problem.
When it comes to law, you're obviously punching about your weight.
I know you're casual on this, but I wanted to chime in to help you learn a thing or two about this (and to do a service to other who might also be oblivious).
Nothing neathack did was illegal, taking it at face value, of course.
Furthermore, he disclosed it responsibly, didn't publish any confidential data, etc,; actually he published no data at all, other than a statistic.
Criminal law deals with intent, it is clear that neathack's intent here was fueled by goodwill and a research purpose. No judge on the planet will look at this and find wrongdoing on his side.
Also read the room, everyone here has told you the same thing ... just say thanks and move on, man.
I care about people downloading stuff they have no right to in the name of security, when actually that wasn't strictly necessary to document the fact that there is an issue.
As soon as you shift from showing that there is a problem to downloading data to establish the size of the problem you are the problem.
Whether he disclosed it responsibly or not is besides the point: you don't get to download a database to prove that it could be done, then you've changed a white hat POC into an actual leak and that is a completely different affair.
FWIW this was in Germany against a server in Sweden if I've understood it all correctly and I would bet a sizeable amount of money that I know more about EU data protection laws than you do.
"What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data."
You can't do that without downloading the data first.
Yes, you cannot determine if there is a breach of meaningless data or privacy data without inspection; as a layman, op appears to have downloaded as little as possible and inspected as little as possible to determine the nature of the problem.
So, let me get this straight: you are advocating that white hat hackers download data in bulk to determine whether or not there is a breach of privacy? Do you realize that at that point in time there is a breach?
I'm really more than a little bit surprised that people here don't seem to know the difference between reporting a vulnerability, which has the potential of being exploited and actually exploiting it.
This phrase is doing a lot of lifting here. Op clearly downloaded far less than is possible.
> Do you realize that at that point in time there is a breach?
As others have said elsewhere, this is a legal opinion and IANAL. Morally: downloading data, not reading it, then deleting it, then reporting it does not strike me as unethical.
Far less than possible but far more than needed to prove there is a vulnerability.
> Morally: downloading data, not reading it, then deleting it, then reporting it does not strike me as unethical.
How will you prove you didn't read it?
How will you prove you deleted it and that you did not make a copy?
The company has no option: it is now documented that there is a breach. They will have to report that breach (which if I were them I'd be doing immediately, they have 48 hours to do so). And then they are going to have to make a tough decision: thank you for your time and effort and hope that you really didn't read it and that you didn't copy it or play it safe and file a case against you.
Lots of companies are going to play it safe, if only because their compliance department is not going to want to take the risk of you popping up a few months later with another interesting analysis of data that you shouldn't have had in the first place. That's what makes this so risky, it turns a clear 'good deed' into something that could well end up in court. Morality rarely factors into such situations.
Leaving aside the legal argument and focusing on the technical side: Lovable is a Swedish company, but websites generated by Lovable are hosted on servers around the world. That said, the actual data is stored on Supabase servers, which -- as far as I know -- are located in the US.
You seem to be confusing what feels intuitively wrong with what is actually unlawful. Law doesn't work on feelings. Where are the terms of use? Do they explicitly prohibit this kind of access? How do they define it?
If a website publishes data and provides a public key that automatically allows a browser to retrieve it, good luck arguing that your "security" was bypassed.
Also, OP mentions confidential information. That doesn't automatically make neathack liable. If I publish someone's medical records in a publicly accessible comment here, are you liable for reading them, or am I liable for publishing them on a website with essentially no access controls?
>an actual leak
Got a link? Or "just trust me, bro"?
>I would bet a sizeable amount of money
Unfortunately, my religion prohibits gambling, which we consider degenerate behavior. But by all means, keep the confidence. It would be even more impressive if it came with an actual legal argument.
He's one of the site's longest-standing users, but his arguments aren't making much sense. I'm sure he can do better, but he's too stubborn to let this go.
I have seen this with a lot of old-timers. I wish this site were a little more relaxed in general.
Ah, true for some reason this had disappeared from my view:
"What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data. Using a confidence threshold of 85%, we found that nearly 4,000 websites leaked such information."
This to me is proof that you actually downloaded the data, you would not have been able to present the conclusions that you did.
Anyway, I wish you best of luck with this whole affair, you did a great job at identifying a vulnerability, and then you turned it into an actual data breach. I hope that lawyer is willing to sign off on you having destroyed the data because that might become a very important issue.
> Nothing neathack did was illegal, taking it at face value, of course.
Law varies significantly between jurisdictions. You're just blankly claiming it's not illegal. People have been convicted for way less than this, in different countries.
Plus, you could be creating more damages to the company by downloading more data like this, and you're not wearing a white hat when you're creating damage. A white-hat is in and out, without creating damage, with a clean report about the issue so the software provider can patch it.
This was supposedly fixed on October 30, but there are reports floating around that for existing projects, the defaults still apply.
Lovable deserves some heat for using this, but this was a choice that was made by Supabase many years ago to support making it easier to onboard potential customers, a deliberate tradeoff of less onboarding pain for less security.
If your goal is to onboard nontechnical users quickly and not have to deal with them churning because they have to deal with thinking about those pesky "security rules" before they can get a working app, it totally makes sense that you would do this, especially if you just drop a get out of jail free "well, it's the end user's fault for not configuring their app securely" in your T&C.
Somehow it's okay to simultaneously set a golden path to less secure software to support your hockey stick growth model and then turn around and blame end users when it all goes wrong.
Any tables created before the defaults put in place are of course still subject to the wide open previous default configuration, because they would risk break existing users apps if they back-ported those.
For example, if you try to access a table named "secret" that doesn’t exist, Supabase might return a hint saying, "Perhaps you meant the table 'public.super_secret'."
Thanks for volunteering that information, I guess.
The amount of misconfigured supabase out there is astounding. And I'm not just talking about your mom's friend who wants a mahjong social network site. Companies backed by sand hill road VCs to the tunes of 10s of millions of dollars expose keys, don't have rls, set anon envs that are really full admin keys, don't rotate keys, etc.. the list goes on. It's wild.
I can confirm that no advanced hacking techniques were used, and anyone with a moderate understanding of how the web works will be able to reproduce the results.
The initial idea behind the analysis was to start with the most basic setup and then move on to more sophisticated attacks. We never reached that second part, as so much sensitive data was already out in the open.
The bar being vastly lowered is the qualitative difference that some in the pro-AI crowd don't seem to appreciate. At their root, many of the issues we currently see with AI are really cultural/procedural issues that have existed in the industry for a long time, but AI has dumped napalm on the dumpster fire and the neighborhood is now catching fire.
These are canaries in the colemines. Platforms need better defaults, deny-by-default DBs, least-priviledge API, automated auth tests, etc. Basically make insecure deployments difficult with AI generated code.
It's incredible how little understanding the average person has about data, privacy, and security.
Allowing anyone to build and host a public app without a basic understanding of these things is like letting people fly a passenger plane without a license.
I strongly disagree with the second part. Building public app should not require same level of skill as flying an airplane. Not even the same as getting a driver license. It had to be simple, secure and compliant.
The problem are product managers and engineers who build things like Lovable, that are not secure by design. Those products have no right to exist, but their alternatives could exist and let their users focus on what matters, instead of paying web development tax.
You’re right, only if the person building the application is using sound judgement, which goes against the premise.
You and I wouldn’t build applications naively that could lead to people dying. People who don’t understand what they’re building might.
People are vibe coding at your hospital. They’re vibe coding at your utility company. They’re vibe coding at your pharma lab. I’ve met these people. They’re smart. They have phds. I tell them about SSL.
I’ve been mentioning it for years, but data, privacy, and security do not matter to most people, despite them saying so. Personally I operate assuming that my IDs, passport and etc. have already been leaked at some point. So, if someone says my PII has leaked once again, I’d just shrug and move on.
When Equifax thing happened, there was basically no repercussions. And that’s the only data breach I recall being in the news cycle for more than a week. Everything else has just been a blip since then. Also as there were no real repercussions other than some fines, companies took it as a signal that nobody cares (rightfully).
Allowing anyone to design, install, and maintain the plumbing, heating, and electricity of your house without a basic understanding of these things is like...
> Allowing anyone to build and host a public app without a basic understanding of these things is like letting people fly a passenger plane without a license
Is it now? How many hundreds of people have died as a result? How much infrastructure was destroyed?
I get the point you're trying to make, but computer security as a field has already a big problem with maximalist mindset, and it doesn't help in keeping systems of consequence actually secure and people safe from danger.
Railway got hit by pretty much this a few days ago. Attacker likely used abliterated models, stolen keys, ~$740k in AI charges. If it happens to them, random Lovable apps are easy targets.
How is that the same thing. Lovable didnt get hacked.. Also hint, there's nothing scary about ablit models, they dont increase someones capability to hack something with AI just reduce annoying refusals, purely ergonomic.
Seeing as this is reported by a German organisation about a Swedish company: has this been filed with the authorities under the GDPR?
Any details on whether (national) data protection authorities have picked it up and warned the companies and people in violation?
The analysis was conducted with the assistance of a lawyer specializing in data protection. A deliberate decision was made to share the dataset with Lovable so they could contact their customers and have the database access issues resolved.
One of the golden rule of webdev is "never build your own auth" for a reason.
To me, it feels like Lovable is an irresponsible product selling the idea that web development is trivial now with LLMs, where the reality is that although it's true that agentic coding means while anybody can spin up a website in like 10 minutes, a professional software engineer who knows what they are doing can still utilize LLMs much more effectively than an amateur ever could.
I don't see any indication in the post that these cases involved someone writing their own authentication code. This is about authorization mistakes, or the lack of any authorization. And that part you have to write yourself in some way (even if you use libraries), someone has to define the rules about which data is accessible under which rules. Using a pre-made library won't save you from defining those rules wrong.
To me only hard rule like this is to never implement your own cryptographic code. A decent developer can easily implement simple session-based authentication safely, or at least I'd suspect the chance of mistakes is no higher than this developer misconfiguring a much more complex library for authentication.
>In its marketing, Lovable repeatedly emphasizes that no technical knowledge is needed to build projects. This can give the impression that the platform takes care of security too. Yet Lovable’s Terms & Conditions place responsibility for applications and data handling on the customers themselves.
My point is that the way Lovable websites are generated pretty much made no auth or mistake prone from scratch auth stack the default path with no regard to the security problem down the line, and the people they are marketing to are not decent developers or developers at all really, but people who doesn't have much knowledge about software, which is irresponsible.
I've started putting OIDC auth in front of my app (at the gateway level) so that no matter how badly my LLM botches the app, requests can't get through.
This only really works with 1-person apps though. If you need multiple users then you have to match the login with specific database records to make sure you're only exposing the right stuff.
I guess that is to say, even if your authentication is great, you can still botch the authorization.
That's not the problem here. Lovable uses the Supabase Auth layer; they just have their own interface on top of it.
Supabase is kind of similar to Firebase in that you usually communicate directly with the browser through a JavaScript lib. So you need to pay close attention to the security rules (RLS policy in Supabase's case) to restrict data per row based on the user auth.
It's easy to get it wrong; that's why so many Firebase and Supabase databases end up exposed. The platforms have improved the UX so that it's not so common anymore, but it's definitely a gun that's easy to shoot yourself with.
You don't have to expose it like that. You can disable that client and use a server to communicate with the database, which reduces the exposed surface in exchange for complexity (by even having a backend at all).
Mix that design with users with zero tech background who know nothing about systems design and LLMs, and you have a recipe for disaster.
We use Lovable as a prototyping tool and then eject it, review, and deploy on a different platform. It works great for that use case.
The thing is that in auth (like in cryptography) the rulebook is 2000 pages and there is an annotation book of another 2000.
If you miss one page, the llm doesn't and will open you like a tuna can.
What 2000 page auth(n) are you referring to? authn may use aspects of cryptography but it’s largely orthogonal. The point is, it’s fine to roll your own authn.
If you don’t want to roll your own authn, you’re delegating to oauth (oidc) identity providers like google/apple/facebook/etc., which not everyone has an account.
No? The rule is don't roll your own crypto. Building your own auth is normal and learning to do so ought to be expected as part of the standard education of a web developer. I suppose next I might hear something like never define your own table schemas...
At least with "lovable", people knew what they were getting into when they sent their PII to an AI slopped project.
But unfortunately, it’s not always this transparent. In the US, doctors (particularly PE owned ones) often use AI in their backend in one form or the other. In one case, I observed their transcription service was using some LLM to help process data into medical record.
In other cases, your data may even be processed/leaked/trained without your explicit consent through a third party vendor that doesn’t interact directly with doctors office (physician office -> order sent to pharmacy -> pharmacy uses LLM to process data).
This is at least the case for the insidious CVS/Caremark conglomerate.
I wonder what will stop or at least force this wave of individual creators to think a bit further. If I leak data at my job it's a massive deal. Like stop working, start calling people and light a cigarette type deal.
Like mentioned elsewhere in this thread, there are doctors vibe coding shite and sending private medical data to Anthropic's servers, probably without even checking the "don't use for training lol" checkbox. Thousands of passports, invoices, whatever, are rn available on the open internet. People posting private company data on PUBLIC github repos (dealt with 2 such cases lol).
These individuals aren't held to the same standards, and right now nobody is enforcing the rules on them.
And maybe that's fine. Maybe we don't want to stifle innovation. But if my passport ends up on the dark web and I get identity theft'd, then I'd be pissed. There are REAL consequences.
Each psychotic vibecoder making 50 slop projects on an account. Also probably counting demo projects. Idk, if I had to bet against any company it would be lovable.
But they have tons of VC money so they'll get aquired by someone in the incestuous circle of SV investors.
Folks are blaming AI, and that's certainly part of it (folks without knowledge and experience can now build polished apps, though there were still plenty of "WTF" security breaches before AI), but I feel like the only thing that can fix this are real, stiff liability consequences.
The defects listed in this report didn't require some complicated 0-day exploit chain - they were just sensitive data stored without any authentication at all, in a database that anyone with Internet access can view. In other words, it would be the equivalent of a company like Public Storage leaving all of its storage units with no locks on them at all.
Affected users should be able to sue for a lot in these basic security failures. And obviously for many of these small websites, the potential damages would quickly bankrupt the site owners, but that's the point. People would think a lot more carefully about how they protect sensitive data if there were real negative consequences for their failures.
I agree that AI is not necessarily at the core of the problem.
To me, the issue is Lovable’s borderline reckless marketing: repeatedly claiming that “everyone” can build a website “without any technical knowledge” (verbatim from their website), while disclaiming all liability for bugs in their terms and placing responsibility for security on their customers.
Customers who hear day in, day out that “software is solved” and buy into Lovable’s narrative may simply have no idea that they could face substantial fines for privacy violations.
EU citizens can sue. It's not easy in every EU country, and ot differs whether you will be compensated.
But in all cases, under GDPR, the one having the data breach is required to act. If they fail, there's steep fines, often proportional to your revenue. It will hurt the business.
IANAL, but I've had several hosting companies. And while most situations where a customers' site is hacked and/or their data leaked, is clear and not the hosters' responsibility. That hoster does have an obligation to provide good tooling (i.e. ssh, not telnet, sftp, not ftp) and to promote good behaviour (i.e. not allowing the backup system to write the backups to public hosted directories). Or else can be held (partially) responsible for breaches.
Lovable has a responsibility, and, as you pointed out in your article, their proposition of "building apps without technical knowledge" (my words), might make theirs to be more than just putting lines in a EULA to waver that responsibility.
Or just the natural consequence of adopting "move fast and break things" from top to bottom for 20 years, a house-of-cards methodology that was destined to fall in on itself whether or not such models ever arrived.
Formal verification doesn't solve all the problems, but alongside concerns like correctness, restrained surface area, supply chain awareness, treating PII as toxic waste to avoid rathee than treasure to collect can leave projects/orgs in a much better place than they let themselves get too.
This is one of the reasons anthropic nerfs their models. They want you to have an enterprise contract with some spend before they’ll allow you a couple of seats on cyber models, which you need since attackers are much more nimble with ablated open weights models.
As a side note to the "nerfing", I worked with a journalist on the analysis, and she wasn’t comfortable running `curl` on the command line. So she asked the AI to do it -- but it refused, citing concerns about cyber abuse.
On the one hand, that’s kind of encouraging. On the other, it feels like the unskippable “don’t copy this movie” warning on a legally purchased DVD -- the one that gets removed from pirated versions.
I can't state my disdain enough for the rabid, careless advancement of AI, and all the bros that keep telling me to catch up or fall behind, while they are actively causing harm to the hundreds of thousands, or millions of users they have, without care or consideration.
Nearly two decades of experience learning how to build safe, performant software, to be replaced by this shit.
We don't care about people, our customers anymore. We actively harm them for profit. This is such a garbage industry now.
Bad take. That's impossible in our society. I wouldn't get updates from my child's kindergarten. I wouldn't be able to use public transit. I can't access my bank. I can't communicate with my family, friends, spouse, work network.
We don't have an option. Which is why a lack of care for users is abhorrent.
Keeping _more_ of your data private is not impossible. It won't necessarily be the most comfortable thing, as tradeoffs will be made, but you can claw back quite a bit of privacy if that is important to you.
Of course, but that's not what the person I was responding to said. They blamed people for trusting services (many of which) they don't have an option to opt out of.
And this whole thread is just proving my point by blaming people and customers for the mistakes of technology companies.
I am interviewing with a small company that has adopted agentic development in a big way and is looking for a vp of engineering to guide things. I tried to present a balanced view of excitement for the possibilities built on first solving governance, guardrails and mitigating the lethal trifecta. Senior executives are not interested and my position likely disqualified me, even after I talked about needing to speak to your client data governance teams and CISOs. Crickets. They actually said we're looking for someone who's a big AI booster. Bullet dodged.
My perspective is to figure out how to shape the way we use AI to show the right way to use it. This comes from a place of caring about the software and its users.
I strongly believe that when used properly, AI can remove a lot of overhead that comes with writing the code (from both a critical thinking standpoint and physically it actually takes a lot more time than people realize to manually type code). You can focus on the design and have your team much more integrated in the designs and specifications of things without it being a slog to document the damn thing.
If I can go off and write up a draft and get feedback from my team in a day instead of in a week, I think that's fine. We can collectively walk through and use everyone's domain knowledge to correct details. Once everyone is onboard with the spec, I can feed that into a model to progressively do small changes rapidly but in such a way that's still manageable and reviewable for myself and the rest of my team.
If I go off and merge 45000 lines of code into main: fucking fire me and my whole team that approved that.
As with all new things that have extreme views on either side (AI is complete shit versus AI is the second coming of Christ in this case), there's generally a middle ground in between that we can collectively correct toward.
The idea that companies invested in "safe, performant software" before AI is laughable. Tech companies have never cared about the customers beyond their effect on KPIs.
Companies never cared, and most engineers didn't care much either, at least not directly. But a lot still cared to cover their ass and try to do things correctly the first time to reduce future work and headaches. Now that it's all just a few prompts away "Who cares? Ship it now, if it's broken we'll just have Claude fix it later."
Sure, but the parent said "this is such a garbage industry now", implying it was not this way before LLMs. The individuals who care about quality can still pursue quality. In fact, they are better equipped to create quality software than ever. It's a matter of spending time and attention.
The underlying problem is that quality isn't profitable. It goes beyond tech. The companies that cared about quality had their lunch eaten by the companies that chased razor-thin margins and produced disposable junk. I'm not sure what it is about AI that makes everyone forget what the world was like five years ago.
While better equipped for security with frontier models (and I agree with that) I don't know any engineers in my circle that have the time to utilise it. Half of them are laid off, by the way.
The rest, including me, are supposed to take up the slack with ridiculous deadlines and pressure. I don't have time anymore. I have to trust that my prompt and testing descriptions and harness will take care of it. And I have no clue whether that's truly happening, because I have another ticket before I get time to review, or send some swarm of test/probe agents at it. And right now I don't trust it. (Literally look at the parent article?)
I did have that time and capacity in the past. Maybe you didn't. But the landscape is, to me, objectively different now.
I fully agree that tech workers have never been less empowered. When I said that individuals have never been better equipped to pursue quality, I meant within projects they control. There will likely never be agency again at a large company.
We're on the same page about most of this, but my perspective differs in a couple of ways:
- I have worked at companies that didn't prioritize quality and had unreasonable deadlines for most of my career. It's the default state of corporations with leverage over their workers.
- This was getting worse before AI. Big tech was the anomaly in terms of pay, perks, benefits, etc, and they were already jointly deciding not to do that anymore around the time of Musk's Twitter layoffs. The main reason they had offered such a compelling environment was to prevent talented people from being hired by their competition, in which case they were perfectly happy to let them "play around" and focus on quality.
I find it frustrating when people imply that employment was an idyllic wonderland before AI ruined everything. AI is another step, potentially the final step, in boiling the middle class. It didn't invent poor working conditions, it brought tech workers down closer to everyone else's level.
The best move right now is to move to a small startup or start something on your own. I fully understand that this is easier said than done, when you have bills to pay and deadlines keep you too exhausted to escape. To some extent, that's probably part of why the deadlines exist.
Exactly. I cared, and the amazing engineers that taught me cared. We didn't have KPIs in 2008, we had code reviews and mean senior engineers (that were secretly kind) that would tear your code apart and send you back until you got it right.
And we were allowed to? I raised so many tickets over time that had no real business impact beyond "this is a security risk" and nearly all of them I was allowed to adress.
Maybe that makes me some kind of tech-boomer, like people responding to me are implying, but I don't really give a shit.
>I can't state my disdain enough for the rabid, careless advancement of AI, and all the bros that keep telling me to catch up or fall behind, while they are actively causing harm to the hundreds of thousands, or millions of users they have, without care or consideration.
Have you ever considered that the people who are pro-AI are not a monolith, and that it can be simultaneously true that you should absolutely be learning how to effectively and safely use LLMs in engineering workflows, while it is also true that there are people who are using them very unsafely?
Beig opposed to learning how to use AI correctly because some people use it recklessly is like being opposed to learning AWS in 2015 because some people left sensitive data in open S3 buckets.
Do not confuse negligent misuse of the tool for the value proposition of the tool itself.
Re-read my post before quoting it and getting defensive. I'm not anti-AI, I use Claude daily, partly because that's all the industry is now, partly because it can be an amazing tool.
I'm furious about slop-shovelling profit-driven development that actively causes harm to people. And the lack of care for, or outright disdain of customers, specifically fuelled by a misuse of AI, and the people telling me to get used to this, the people telling me to stop reading my own PRs, the people telling me this is just how it is now. When we all know that doesn't have to be true.
Just bide your time until there are some absolutely catastrophic breaches like what is happening in Japan right now on a daily basis, and then people and legislators will slowly start demanding accountability.
I just don't see any grassroots movement advocating for responsible software development winning out with a demographic that can't even unionize for their own self benefit.
There is no part of me that believes people will fundamentally do the right thing without an incentive to do so (money) or punishment (fines, jail). There won't be action until there is so much suffering that governments can no longer ignore it.
It's surprising how different the standards are between software engineering and all other kinds of engineering, for instance civil or mechanic.
The old joke was "If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization", but in the AI era it feels like a butterfly would be enough.
Counterpoint: if programmers wrote programs the way structural engineers designed buildings, ecommerce apps would take half a decade to build and the software industry would have just gotten around to inventing Windows Vista and jQuery in the last year or so.
If you're saying e-commerce would simply be consistent and secure product catalogs with fewer promotional and tracking dark patterns, web scripting would rely on a and extensible and functional library for manipulating markup documents, and a stable OS kernel with few dark patterns and no cloud accounts, many of us would be sold in a heartbeat.
I don't think you're making the counterpoint you mean to be.
Don't be naïve - the industry wouldn't be any less laden with surveillance crapitalism; the presence of that isn't a product of the pace of development, it is a product of the nature of consumers to eagerly suffer limitless abuse so long as the product "just werks", looks/feels polished, and doesn't require any kind of learning or discomfort.
Freedom from everything you enumerated already exists for anyone who cares enough to pursue it - robust ad and script blockers, operating systems with no coercion or user hostility, simple and stable scripting languages. They're not as polished as the products of manipulative and extractive abusers like Microsoft, Google, Apple, Discord, et al., and accordingly, users shun them, for the reason I explained above.
> it is a product of the nature of consumers to eagerly suffer limitless abuse
Ah, there's the victim blaming.
> Freedom from everything you enumerated already exists for anyone who cares enough to pursue it
Do you have any idea how daily life is for non-technical people? It isn't reasonable in our society to expect everyone to switch to FOSS, and blaming 60 year olds for getting their identity stolen because they weren't using a Linux phone or rolling their own self hosted matrix chat server is such a cruel and unbelievably out of touch take.
Hey, if the option is on the table for states to fight back, up and including enacting the corporate death penalty (forced dissolution) against Microsoft, Alphabet, Meta, Apple, etc, I'm all for that.
That would never happen though, the victims would plea and fight too vigorously in defense of their own abusers.
Who should be blamed for the actions of corporations ultimately subject to the society-scale approval of their consumers? No amount of misbehavior, malice, negligence, or abuse from the company can alienate the modal user of their products, they've thoroughly demonstrated this too many times to count.
It's not the doddering elderly folks I blame the most, it's everyone younger than them, the folks who grew up with the technology, aware of every abuse by these companies, who refuse to change their behavior. The ones who know better. The ones who can explain to you in great detail all about the surveillance, the dark patterns, the emotional manipulation, the extractive commercial angle, and who all refuse to adjust their behavior anyway. It's a choice, and it's one they make on a continual basis.
And to answer your question, no, I don't have much of an idea how daily life is for non-technical people besides them voluntarily suffering abuse like theft of their time (advertisements), or voluntarily paying a bribe to avoid the abuse, for no real reason other than that they don't care enough about the abuse or the bribe to spend 5 minutes learning how avoid both for free for the rest of their life.
1. A desire to harm others or to see others suffer; extreme ill will or spite.
2. The intent to commit an unlawful act without justification or excuse.
3. An improper motive for an action, such as desire to cause injury to another.
None of the above apply to me. I don't want to harm or injure anyone, and I don't want anyone to suffer. Consumers choose suffering and choose to voluntarily accept harm and injury, and I wish they wouldn't.
But you can't force people to not harm themselves, it's something they have to choose for themselves. Anyone who's ever lived with or loved an alcoholic or drug addict understands this. The contempt isn't for the person, it's for the theft of the person's faculties. The response to the person being unwilling to change is frustration with their choices, not contempt for the human being.
Hi, I’m the author of this blog post. I submitted it to HN two weeks ago, but it didn’t gain much traction, so I’m glad to see it resubmitted. Happy to answer any questions, though I think the situation is fairly self-explanatory.
How many of the 159k analyzed sites leaked private data?
Edit: The answer (as neathack tells it) is: 2.5% of the sites. The reply was buried below.
From the blog post:
What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data. Using a confidence threshold of 85%, we found that nearly 4,000 websites leaked such information.Edit: That's 15,9% of all Lovable websites that use Supabase.
Don't you think you should have reported this after you established the principle rather than to download more than you needed to do so? Because it looks to me as though you went from 'white hat hacker' to 'quite probably criminal' without a really good reason.
Do you think Lovable would be transparent as to the extent of the problem? Or would they use terminology like “a small subset of our users” and “data not including card details” or whatever to minimize the issue.
I’d much rather live in a world where independent researchers can figure out the extent of the issue rather than go “uh oh there might be an issue here, guess we’ll never know the extent of it, let’s trust them to fix it”.
That's a good question. But it is a very fine line between being a white hat that reports a vulnerability in good faith and someone who actually exploits that vulnerability and turns a vulnerability report into an actual breach. This is - in my opinion, clearly the HN hive mind disagrees - across that line.
For instance, if I run into a server and I see 'passport.jpg' in a directory listing I could download it to see if it really is a passport. But at that point in time I am flirting with becoming a defendant in a data breach rather than that I'm reporting a possible vulnerability. Sure, you could argue that the site owner might close the hole and then they could downplay that it really was a passport photograph. But I care more about the hole being closed than that I care about the contents of the file, which clearly are not meant to be disclosed based on the filename alone.
Independent researchers that are under contract with the companies they review sometimes have more leeway in what they are allowed to do. But downloading enough data to be able to run statistical analysis using LLMs is turning this into a reportable breach and that is a great way to get this whole thing to backfire, possibly to the point where the OP turns defendant.
One thought might be - did they
the article mentions "we retrieved only parts of the datasets rather than downloading them in full", so a charitable interpretation might be that for each identified source candidate, they spot checked (or had a model spot check) a few individual records to have incontrovertible proof of breach of privacy, and then ran an aggregate count query to gauge the extent of breach. I think that would be fairly in line with "white hat" behavior that's doing the smallest amount of extra work towards a solid proof and clearly not taking the obvious and easy next step of downloading the remainder that would be required for self enrichment. Difficult to argue that there is an intention of malice.
A bad actor would be downloading /all/ the data and disclosing it to no one, and selling it while obscuring the source.
That said I agree that that certainly might fall afoul of the letter of whatever local laws apply, and certainly prosecutors can also be over-zealous in their interpretations. So personally I would never risk doing something that comes up against that line, from a personal risk POV. But idk if it's fair to argue that this is NOT good faith.
I don't argue it is not good faith. I just wonder if OP realized that they might be crossing a line they shouldn't be crossing, ever, while doing security research without being under contract to the party they are investigating.
Then at least you are under and NDA and there is no way for the company to spin it that you were a bad guy.
The numbers mentioned are pretty serious: "We found records containing email addresses in 10,560 tables across 7,403 of the 24,664 websites (30%)."
To me - and please correct me if I'm wrong - that reads like 'we downloaded at least parts of 10560 tables'.
How do you know which websites are affected without sampling data from each website in question? As the blog explains, it's safe to serve public data from Supabase in this manner, so the only way to determine whether the website has an issue is to look at the actual records they are serving. Their approach seems entirely reasonable to me.
As someone who has dealt with vulnerability reports on products before, I'd strongly prefer getting concrete proof of an issue versus "your product came back as positive on our vulnerability scanner^tm" which is 99% of the time going to be a scam or a waste of time.
Yes, would you like the report to be what you need to know or a report that will cause you to have to report an actual breach rather than just a vulnerability? Because that is what this is.
How does that relate to e.g. have I been pwnd? If the obligation to be a white hat is to protect the owners of the information then is it not also to inform the persons who have been impacted a valid white hat activity? From my perspective the classification of someone as a black hat should be if they use it for something malicious. And if anything disclosure limits any potential use. On the other hand our big LLM companies Hoover over everything, have access to it, occasionally go rogue with their knowledge and access, but I don’t see many people calling them black hats.
You'll notice that the person that built have I been pwnd has been extremely careful in how the data comes into their possession, I highly doubt they would ever go so far as to breach a site themselves. It would go completely against their mission to do so.
> On the other hand our big LLM companies Hoover over everything, have access to it, occasionally go rogue with their knowledge and access, but I don’t see many people calling them black hats.
I'll be more than happy to call them criminals and I've warned very publicly about their resetting configuration switches to allow them to pretend that they have your consent to use your data for training purposes.
I should also mention that the analysis was conducted in collaboration with a lawyer specializing in data protection and ethical hacking.
That's an important bit of information that probably should make it into the blog, and I hope you guys made the right call, but I think that that is probably not the case. Best of luck.
oh: "What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data."
I hope at least that that was a local model. If not you've made things considerably worse.
It’s mentioned in the Handelsblatt article linked in the blog post. Since the lawyer was brought in by the newspaper, we didn’t mention their involvement in the post.
All LLM-based analysis was performed using local models, specifically the "elephant-1" model on our local RTX Pro 4000 SFF.
I respect your insistence on proper data handling -- you’re right in doing so -- and we tried to be as careful as possible when conducting the analysis.
Now let’s hold Lovable to the same standard.
The lawyer that was brought in by the newspaper was making sure the newspaper wouldn't end up being liable. The only lawyer whose advice you should take is the one that you pay.
As long as they are representing someone I think lawyers have the same obligations regardless of who actually pays them or if they are receiving any payment at all.
Yes, this is true, but that requires there to not be any conflict of interest, which is really not all that clear-cut in situations like these.
Need to know the full blast radius and severity...
What do you mean? This has been disclosed to, and the full dataset has been shared with, Lovable.
What I mean is: you knew it worked before you went and pulled more data, you could have stopped after the first two records, why pull more than you needed to prove that it was possible?
We retrieved only enough data to determine whether it was privacy-sensitive, which could not be reliably assessed from just two records. We did not download any more than that.
Supabase tables also use pagination, and we never downloaded more than the first page.
I may not be the intended audience, but as the author of a similar system having a benchmark like this of what user behavior can result in when it’s unconstrained is helpful for my safety evals. So in an indirect way, you’re making future platforms more secure by sharing this.
That's great to hear!
You probably should not have done that. It's one thing to determine that a hole exists, an entirely different one to exploit it to the point that you can do damage assessment on the target.
Would you prefer someone that didn't intent to report but to exploit the data downloaded everything and the company and you knew fuck all about it?
That's not the only choice available.
why don't you chill the fuck out. you've been harassing the op over and over here
I think you should take your own advice.
Well, it seems that no amount of explanations will satisfy you, so I would say you're the one pushing for a binary choice scenario. Hard...
So I can, maybe, understand your posture: what are your ties to Lovable?
None whatsoever. What kind of a dumb question is that?
You seem to be insanely hellbent on implying the actually worst possible intent to the researcher who acted in the good faith and contacted the vendor aiming for the quickest and best fix.
It looks like you're weirdly trying to inflate a legitimate attempt to confirm if the severity of the exposure is as bad as suspected into a full blown nasty data thief.
This is why everyone is so puzzled. You're genuinely harassing op for doing a good thing and responsibly reporting the severe security problem.
I haven't made a single statement about their intent.
I have just pointed out that there is a difference between reporting a vulnerability and actually exploiting that vulnerability.
the concern trolling discourse of hacker news strikes again
When it comes to law, you're obviously punching about your weight.
I know you're casual on this, but I wanted to chime in to help you learn a thing or two about this (and to do a service to other who might also be oblivious).
Nothing neathack did was illegal, taking it at face value, of course.
Furthermore, he disclosed it responsibly, didn't publish any confidential data, etc,; actually he published no data at all, other than a statistic.
Criminal law deals with intent, it is clear that neathack's intent here was fueled by goodwill and a research purpose. No judge on the planet will look at this and find wrongdoing on his side.
Also read the room, everyone here has told you the same thing ... just say thanks and move on, man.
I don't care about the room.
I care about people downloading stuff they have no right to in the name of security, when actually that wasn't strictly necessary to document the fact that there is an issue.
As soon as you shift from showing that there is a problem to downloading data to establish the size of the problem you are the problem.
Whether he disclosed it responsibly or not is besides the point: you don't get to download a database to prove that it could be done, then you've changed a white hat POC into an actual leak and that is a completely different affair.
FWIW this was in Germany against a server in Sweden if I've understood it all correctly and I would bet a sizeable amount of money that I know more about EU data protection laws than you do.
> As soon as you shift from showing that there is a problem to downloading data to establish the size of the problem
This is not a self-evident argument, what are you basing this on?
Problem scope is part of problem definition. If you dont know the scope, you dont know the problem.
On this quote by the OP:
"What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data."
You can't do that without downloading the data first.
Yes, you cannot determine if there is a breach of meaningless data or privacy data without inspection; as a layman, op appears to have downloaded as little as possible and inspected as little as possible to determine the nature of the problem.
So, let me get this straight: you are advocating that white hat hackers download data in bulk to determine whether or not there is a breach of privacy? Do you realize that at that point in time there is a breach?
I'm really more than a little bit surprised that people here don't seem to know the difference between reporting a vulnerability, which has the potential of being exploited and actually exploiting it.
> download data in bulk
This phrase is doing a lot of lifting here. Op clearly downloaded far less than is possible.
> Do you realize that at that point in time there is a breach?
As others have said elsewhere, this is a legal opinion and IANAL. Morally: downloading data, not reading it, then deleting it, then reporting it does not strike me as unethical.
Far less than possible but far more than needed to prove there is a vulnerability.
> Morally: downloading data, not reading it, then deleting it, then reporting it does not strike me as unethical.
How will you prove you didn't read it?
How will you prove you deleted it and that you did not make a copy?
The company has no option: it is now documented that there is a breach. They will have to report that breach (which if I were them I'd be doing immediately, they have 48 hours to do so). And then they are going to have to make a tough decision: thank you for your time and effort and hope that you really didn't read it and that you didn't copy it or play it safe and file a case against you.
Lots of companies are going to play it safe, if only because their compliance department is not going to want to take the risk of you popping up a few months later with another interesting analysis of data that you shouldn't have had in the first place. That's what makes this so risky, it turns a clear 'good deed' into something that could well end up in court. Morality rarely factors into such situations.
Leaving aside the legal argument and focusing on the technical side: Lovable is a Swedish company, but websites generated by Lovable are hosted on servers around the world. That said, the actual data is stored on Supabase servers, which -- as far as I know -- are located in the US.
>they have no right to
You seem to be confusing what feels intuitively wrong with what is actually unlawful. Law doesn't work on feelings. Where are the terms of use? Do they explicitly prohibit this kind of access? How do they define it?
If a website publishes data and provides a public key that automatically allows a browser to retrieve it, good luck arguing that your "security" was bypassed.
Also, OP mentions confidential information. That doesn't automatically make neathack liable. If I publish someone's medical records in a publicly accessible comment here, are you liable for reading them, or am I liable for publishing them on a website with essentially no access controls?
>an actual leak
Got a link? Or "just trust me, bro"?
>I would bet a sizeable amount of money
Unfortunately, my religion prohibits gambling, which we consider degenerate behavior. But by all means, keep the confidence. It would be even more impressive if it came with an actual legal argument.
OP has ninja edited their comment to remove proof of exfiltration. Now you are free to call me a liar I guess...
I reread the whole thing and could not find any ninja edits. You are entering the domain of bad faith here.
No, I made a simple mistake. The comment I was referring to was collapsed.
Wait, what did I do? Where did I "remove a proof of exfiltration"? I'm confused.
It's sad to see him tripping like that.
He's one of the site's longest-standing users, but his arguments aren't making much sense. I'm sure he can do better, but he's too stubborn to let this go.
I have seen this with a lot of old-timers. I wish this site were a little more relaxed in general.
Your concern is appreciated :)
Ah, true for some reason this had disappeared from my view:
"What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data. Using a confidence threshold of 85%, we found that nearly 4,000 websites leaked such information."
This to me is proof that you actually downloaded the data, you would not have been able to present the conclusions that you did.
Anyway, I wish you best of luck with this whole affair, you did a great job at identifying a vulnerability, and then you turned it into an actual data breach. I hope that lawyer is willing to sign off on you having destroyed the data because that might become a very important issue.
> Nothing neathack did was illegal, taking it at face value, of course.
Law varies significantly between jurisdictions. You're just blankly claiming it's not illegal. People have been convicted for way less than this, in different countries.
Plus, you could be creating more damages to the company by downloading more data like this, and you're not wearing a white hat when you're creating damage. A white-hat is in and out, without creating damage, with a clean report about the issue so the software provider can patch it.
Spot on.
>by downloading more data like this
What are you talking about? Which data? Is this data in the room with us right now?
You probably should read all that the OP has written before you comment further.
Thanks! So ~2.5%.
The part people are ignoring is Supabase's culpability in this matter: https://github.com/orgs/supabase/discussions/4547
This was supposedly fixed on October 30, but there are reports floating around that for existing projects, the defaults still apply.
Lovable deserves some heat for using this, but this was a choice that was made by Supabase many years ago to support making it easier to onboard potential customers, a deliberate tradeoff of less onboarding pain for less security.
> I've realised that the supabaseKey needed for creating the client on the frontend also grants users access to the database by default.
I can't believe someone actually made this by design
I believe it
If your goal is to onboard nontechnical users quickly and not have to deal with them churning because they have to deal with thinking about those pesky "security rules" before they can get a working app, it totally makes sense that you would do this, especially if you just drop a get out of jail free "well, it's the end user's fault for not configuring their app securely" in your T&C.
Somehow it's okay to simultaneously set a golden path to less secure software to support your hockey stick growth model and then turn around and blame end users when it all goes wrong.
This 5 year old issue was only just fixed this October? Got a link to where it was fixed?
https://github.com/orgs/supabase/discussions/45329
Any tables created before the defaults put in place are of course still subject to the wide open previous default configuration, because they would risk break existing users apps if they back-ported those.
While we’re at it, I was quite surprised to discover that Supabase leaks schema information through the `hint` field in error messages:
https://docs.postgrest.org/en/latest/references/errors.html
For example, if you try to access a table named "secret" that doesn’t exist, Supabase might return a hint saying, "Perhaps you meant the table 'public.super_secret'."
Thanks for volunteering that information, I guess.
nobody needs supabase.
The amount of misconfigured supabase out there is astounding. And I'm not just talking about your mom's friend who wants a mahjong social network site. Companies backed by sand hill road VCs to the tunes of 10s of millions of dollars expose keys, don't have rls, set anon envs that are really full admin keys, don't rotate keys, etc.. the list goes on. It's wild.
One might argue that data breaches been there way before AI, the difference with these AI generated apps is:
- They are an easy target, they don’t need any sophisticated hacking knowledge (article mentions inspecting network calls in browsers).
- There tons of those, and more created everyday. For the average user, they don’t know what they are using, or how risky it can be.
I can confirm that no advanced hacking techniques were used, and anyone with a moderate understanding of how the web works will be able to reproduce the results.
The initial idea behind the analysis was to start with the most basic setup and then move on to more sophisticated attacks. We never reached that second part, as so much sensitive data was already out in the open.
Some don't even need to be hacked. The person just thought they didn't need a login and left everything open.
The bar being vastly lowered is the qualitative difference that some in the pro-AI crowd don't seem to appreciate. At their root, many of the issues we currently see with AI are really cultural/procedural issues that have existed in the industry for a long time, but AI has dumped napalm on the dumpster fire and the neighborhood is now catching fire.
These are canaries in the colemines. Platforms need better defaults, deny-by-default DBs, least-priviledge API, automated auth tests, etc. Basically make insecure deployments difficult with AI generated code.
It's incredible how little understanding the average person has about data, privacy, and security.
Allowing anyone to build and host a public app without a basic understanding of these things is like letting people fly a passenger plane without a license.
I strongly disagree with the second part. Building public app should not require same level of skill as flying an airplane. Not even the same as getting a driver license. It had to be simple, secure and compliant.
The problem are product managers and engineers who build things like Lovable, that are not secure by design. Those products have no right to exist, but their alternatives could exist and let their users focus on what matters, instead of paying web development tax.
Those are not even remotely equivalent costs.
You’re right, only if the person building the application is using sound judgement, which goes against the premise.
You and I wouldn’t build applications naively that could lead to people dying. People who don’t understand what they’re building might.
People are vibe coding at your hospital. They’re vibe coding at your utility company. They’re vibe coding at your pharma lab. I’ve met these people. They’re smart. They have phds. I tell them about SSL.
Nobody is dying from this leak.
I'm sure some people are having their lives turned upside down by these data breaches.
domino effect, these leaked credentials could easily end up on the darknet and used as data for black hat tracking tools which criminals use
> these leaked credentials could easily end up on the darknet and used as data for black hat tracking tools which criminals use
And then they could do something else that might result in property damage. Not at all comparable to the danger of an unlicensed pilot.
I’ve been mentioning it for years, but data, privacy, and security do not matter to most people, despite them saying so. Personally I operate assuming that my IDs, passport and etc. have already been leaked at some point. So, if someone says my PII has leaked once again, I’d just shrug and move on.
When Equifax thing happened, there was basically no repercussions. And that’s the only data breach I recall being in the news cycle for more than a week. Everything else has just been a blip since then. Also as there were no real repercussions other than some fines, companies took it as a signal that nobody cares (rightfully).
It's incredible how little understanding the average person has about plumbing, heating, and electricity.
And I do not try to “vibe” plumb or do the electricity or heating of my house.
They're also all licensed professions.
only in some jurisdictions. in many places the job titles are not even protected.
Allowing anyone to design, install, and maintain the plumbing, heating, and electricity of your house without a basic understanding of these things is like...
> Allowing anyone to build and host a public app without a basic understanding of these things is like letting people fly a passenger plane without a license
Is it now? How many hundreds of people have died as a result? How much infrastructure was destroyed?
I get the point you're trying to make, but computer security as a field has already a big problem with maximalist mindset, and it doesn't help in keeping systems of consequence actually secure and people safe from danger.
careful what you wish for. do you want outlaw teenagers of playing with computers until they get government issued web license?
I could imagine licensing for handling certain forms of data. Or at the very least, liability.
If they are running a company handling sensitive customer data…then yes.
If that user is European the answer is almost definitely a 100% serious yes.
After dryers and AC, the EU is coming for your websites!!!
Railway got hit by pretty much this a few days ago. Attacker likely used abliterated models, stolen keys, ~$740k in AI charges. If it happens to them, random Lovable apps are easy targets.
How is that the same thing. Lovable didnt get hacked.. Also hint, there's nothing scary about ablit models, they dont increase someones capability to hack something with AI just reduce annoying refusals, purely ergonomic.
Even less incentive for people to pay startups
How can they make such a mistake? Surely theyre such a big company, like how does this slip by?
Seeing as this is reported by a German organisation about a Swedish company: has this been filed with the authorities under the GDPR? Any details on whether (national) data protection authorities have picked it up and warned the companies and people in violation?
No, it has not been reported to the authorities.
The analysis was conducted with the assistance of a lawyer specializing in data protection. A deliberate decision was made to share the dataset with Lovable so they could contact their customers and have the database access issues resolved.
In your opinion, what would allow the intentions behind GDPR to be better realized?
One of the golden rule of webdev is "never build your own auth" for a reason.
To me, it feels like Lovable is an irresponsible product selling the idea that web development is trivial now with LLMs, where the reality is that although it's true that agentic coding means while anybody can spin up a website in like 10 minutes, a professional software engineer who knows what they are doing can still utilize LLMs much more effectively than an amateur ever could.
I don't see any indication in the post that these cases involved someone writing their own authentication code. This is about authorization mistakes, or the lack of any authorization. And that part you have to write yourself in some way (even if you use libraries), someone has to define the rules about which data is accessible under which rules. Using a pre-made library won't save you from defining those rules wrong.
To me only hard rule like this is to never implement your own cryptographic code. A decent developer can easily implement simple session-based authentication safely, or at least I'd suspect the chance of mistakes is no higher than this developer misconfiguring a much more complex library for authentication.
In the article:
>In its marketing, Lovable repeatedly emphasizes that no technical knowledge is needed to build projects. This can give the impression that the platform takes care of security too. Yet Lovable’s Terms & Conditions place responsibility for applications and data handling on the customers themselves.
My point is that the way Lovable websites are generated pretty much made no auth or mistake prone from scratch auth stack the default path with no regard to the security problem down the line, and the people they are marketing to are not decent developers or developers at all really, but people who doesn't have much knowledge about software, which is irresponsible.
I've started putting OIDC auth in front of my app (at the gateway level) so that no matter how badly my LLM botches the app, requests can't get through.
This only really works with 1-person apps though. If you need multiple users then you have to match the login with specific database records to make sure you're only exposing the right stuff.
I guess that is to say, even if your authentication is great, you can still botch the authorization.
That's not the problem here. Lovable uses the Supabase Auth layer; they just have their own interface on top of it.
Supabase is kind of similar to Firebase in that you usually communicate directly with the browser through a JavaScript lib. So you need to pay close attention to the security rules (RLS policy in Supabase's case) to restrict data per row based on the user auth.
It's easy to get it wrong; that's why so many Firebase and Supabase databases end up exposed. The platforms have improved the UX so that it's not so common anymore, but it's definitely a gun that's easy to shoot yourself with.
You don't have to expose it like that. You can disable that client and use a server to communicate with the database, which reduces the exposed surface in exchange for complexity (by even having a backend at all).
Mix that design with users with zero tech background who know nothing about systems design and LLMs, and you have a recipe for disaster.
We use Lovable as a prototyping tool and then eject it, review, and deploy on a different platform. It works great for that use case.
Where is that rule written? You can build a really solid auth system, if you follow the rules. Also, what kind of auth method are you mentioning?
The thing is that in auth (like in cryptography) the rulebook is 2000 pages and there is an annotation book of another 2000. If you miss one page, the llm doesn't and will open you like a tuna can.
What 2000 page auth(n) are you referring to? authn may use aspects of cryptography but it’s largely orthogonal. The point is, it’s fine to roll your own authn.
If you don’t want to roll your own authn, you’re delegating to oauth (oidc) identity providers like google/apple/facebook/etc., which not everyone has an account.
No? The rule is don't roll your own crypto. Building your own auth is normal and learning to do so ought to be expected as part of the standard education of a web developer. I suppose next I might hear something like never define your own table schemas...
At least with "lovable", people knew what they were getting into when they sent their PII to an AI slopped project.
But unfortunately, it’s not always this transparent. In the US, doctors (particularly PE owned ones) often use AI in their backend in one form or the other. In one case, I observed their transcription service was using some LLM to help process data into medical record.
In other cases, your data may even be processed/leaked/trained without your explicit consent through a third party vendor that doesn’t interact directly with doctors office (physician office -> order sent to pharmacy -> pharmacy uses LLM to process data).
This is at least the case for the insidious CVS/Caremark conglomerate.
> At least with "lovable", people knew what they were getting into when they sent their PII to an AI slopped project.
The people sharing their PII usually don't know that it's Lovable.
i think this is old news tho.
I wonder what will stop or at least force this wave of individual creators to think a bit further. If I leak data at my job it's a massive deal. Like stop working, start calling people and light a cigarette type deal.
Like mentioned elsewhere in this thread, there are doctors vibe coding shite and sending private medical data to Anthropic's servers, probably without even checking the "don't use for training lol" checkbox. Thousands of passports, invoices, whatever, are rn available on the open internet. People posting private company data on PUBLIC github repos (dealt with 2 such cases lol).
These individuals aren't held to the same standards, and right now nobody is enforcing the rules on them.
And maybe that's fine. Maybe we don't want to stifle innovation. But if my passport ends up on the dark web and I get identity theft'd, then I'd be pissed. There are REAL consequences.
> 24,664 websites sent database records from Supabase as soon as they loaded, without requiring a login or any other interaction.
Oh boy, open season on those entry points
Who uses lovable lol. Hire someone you can sue if need be to make your websites.
According to Lovable, more than 50 million projects have been created using its platform.
Each psychotic vibecoder making 50 slop projects on an account. Also probably counting demo projects. Idk, if I had to bet against any company it would be lovable.
But they have tons of VC money so they'll get aquired by someone in the incestuous circle of SV investors.
makes me wonder if Retool’s governance focused message is valid
Why do I only ever hear about Supabase in the context of people getting fucked?
I think part of it is because they often just the default answer the LLMS spit out when making these projects.
Folks are blaming AI, and that's certainly part of it (folks without knowledge and experience can now build polished apps, though there were still plenty of "WTF" security breaches before AI), but I feel like the only thing that can fix this are real, stiff liability consequences.
The defects listed in this report didn't require some complicated 0-day exploit chain - they were just sensitive data stored without any authentication at all, in a database that anyone with Internet access can view. In other words, it would be the equivalent of a company like Public Storage leaving all of its storage units with no locks on them at all.
Affected users should be able to sue for a lot in these basic security failures. And obviously for many of these small websites, the potential damages would quickly bankrupt the site owners, but that's the point. People would think a lot more carefully about how they protect sensitive data if there were real negative consequences for their failures.
I agree that AI is not necessarily at the core of the problem.
To me, the issue is Lovable’s borderline reckless marketing: repeatedly claiming that “everyone” can build a website “without any technical knowledge” (verbatim from their website), while disclaiming all liability for bugs in their terms and placing responsibility for security on their customers.
Customers who hear day in, day out that “software is solved” and buy into Lovable’s narrative may simply have no idea that they could face substantial fines for privacy violations.
EU citizens can sue. It's not easy in every EU country, and ot differs whether you will be compensated.
But in all cases, under GDPR, the one having the data breach is required to act. If they fail, there's steep fines, often proportional to your revenue. It will hurt the business.
E.g. here's what a ditch citizen can do and expect https://www.autoriteitpersoonsgegevens.nl/en/themes/security...
The tricky thing is: It's not Lovable that can be sued.
I wouldn't be too sure about that.
IANAL, but I've had several hosting companies. And while most situations where a customers' site is hacked and/or their data leaked, is clear and not the hosters' responsibility. That hoster does have an obligation to provide good tooling (i.e. ssh, not telnet, sftp, not ftp) and to promote good behaviour (i.e. not allowing the backup system to write the backups to public hosted directories). Or else can be held (partially) responsible for breaches.
Lovable has a responsibility, and, as you pointed out in your article, their proposition of "building apps without technical knowledge" (my words), might make theirs to be more than just putting lines in a EULA to waver that responsibility.
Nothing to see here and working as intended.
Imagine frontier abliterated models running at 2,400 TPS finding all these issues rapidly on web apps like this.
You are not ready security-wise when these models get faster. ("Formal verification" won't save you either)
Nothing saves if human is completely out of the loop
Or just the natural consequence of adopting "move fast and break things" from top to bottom for 20 years, a house-of-cards methodology that was destined to fall in on itself whether or not such models ever arrived.
Formal verification doesn't solve all the problems, but alongside concerns like correctness, restrained surface area, supply chain awareness, treating PII as toxic waste to avoid rathee than treasure to collect can leave projects/orgs in a much better place than they let themselves get too.
Long way to go from here though.
This is one of the reasons anthropic nerfs their models. They want you to have an enterprise contract with some spend before they’ll allow you a couple of seats on cyber models, which you need since attackers are much more nimble with ablated open weights models.
As a side note to the "nerfing", I worked with a journalist on the analysis, and she wasn’t comfortable running `curl` on the command line. So she asked the AI to do it -- but it refused, citing concerns about cyber abuse.
On the one hand, that’s kind of encouraging. On the other, it feels like the unskippable “don’t copy this movie” warning on a legally purchased DVD -- the one that gets removed from pirated versions.
It seems to punish the wrong people.
The ultimate capitalist model, create a never before experienced problem and sell the solution on a never ending subscription. Progress!
I can't state my disdain enough for the rabid, careless advancement of AI, and all the bros that keep telling me to catch up or fall behind, while they are actively causing harm to the hundreds of thousands, or millions of users they have, without care or consideration.
Nearly two decades of experience learning how to build safe, performant software, to be replaced by this shit.
We don't care about people, our customers anymore. We actively harm them for profit. This is such a garbage industry now.
Well the customer also shouldn’t trust anyone, especially not with handling their private data.
How can the customer know?
Bad take. That's impossible in our society. I wouldn't get updates from my child's kindergarten. I wouldn't be able to use public transit. I can't access my bank. I can't communicate with my family, friends, spouse, work network.
We don't have an option. Which is why a lack of care for users is abhorrent.
Keeping _more_ of your data private is not impossible. It won't necessarily be the most comfortable thing, as tradeoffs will be made, but you can claw back quite a bit of privacy if that is important to you.
Of course, but that's not what the person I was responding to said. They blamed people for trusting services (many of which) they don't have an option to opt out of.
And this whole thread is just proving my point by blaming people and customers for the mistakes of technology companies.
I am interviewing with a small company that has adopted agentic development in a big way and is looking for a vp of engineering to guide things. I tried to present a balanced view of excitement for the possibilities built on first solving governance, guardrails and mitigating the lethal trifecta. Senior executives are not interested and my position likely disqualified me, even after I talked about needing to speak to your client data governance teams and CISOs. Crickets. They actually said we're looking for someone who's a big AI booster. Bullet dodged.
When the Republic gains in strength, the Jedi go hiding in the desert.
A lot of CTOs should be looking into D&O insurance.
They sound like a team looking to get funding or sell out and need the AI sauce to sweeten the deal, not deal with messy real business needs.
My perspective is to figure out how to shape the way we use AI to show the right way to use it. This comes from a place of caring about the software and its users.
I strongly believe that when used properly, AI can remove a lot of overhead that comes with writing the code (from both a critical thinking standpoint and physically it actually takes a lot more time than people realize to manually type code). You can focus on the design and have your team much more integrated in the designs and specifications of things without it being a slog to document the damn thing.
If I can go off and write up a draft and get feedback from my team in a day instead of in a week, I think that's fine. We can collectively walk through and use everyone's domain knowledge to correct details. Once everyone is onboard with the spec, I can feed that into a model to progressively do small changes rapidly but in such a way that's still manageable and reviewable for myself and the rest of my team.
If I go off and merge 45000 lines of code into main: fucking fire me and my whole team that approved that.
As with all new things that have extreme views on either side (AI is complete shit versus AI is the second coming of Christ in this case), there's generally a middle ground in between that we can collectively correct toward.
The idea that companies invested in "safe, performant software" before AI is laughable. Tech companies have never cared about the customers beyond their effect on KPIs.
Companies never cared, and most engineers didn't care much either, at least not directly. But a lot still cared to cover their ass and try to do things correctly the first time to reduce future work and headaches. Now that it's all just a few prompts away "Who cares? Ship it now, if it's broken we'll just have Claude fix it later."
In the past the price of caring was that you were slightly slower, but it would be obvious that the delivered quality was better.
Now you'll be at least 50% slower and likely fired if you care, because caring means understanding the code.
But people care. A self respecting engineer cares.
Sure, but the parent said "this is such a garbage industry now", implying it was not this way before LLMs. The individuals who care about quality can still pursue quality. In fact, they are better equipped to create quality software than ever. It's a matter of spending time and attention.
The underlying problem is that quality isn't profitable. It goes beyond tech. The companies that cared about quality had their lunch eaten by the companies that chased razor-thin margins and produced disposable junk. I'm not sure what it is about AI that makes everyone forget what the world was like five years ago.
While better equipped for security with frontier models (and I agree with that) I don't know any engineers in my circle that have the time to utilise it. Half of them are laid off, by the way.
The rest, including me, are supposed to take up the slack with ridiculous deadlines and pressure. I don't have time anymore. I have to trust that my prompt and testing descriptions and harness will take care of it. And I have no clue whether that's truly happening, because I have another ticket before I get time to review, or send some swarm of test/probe agents at it. And right now I don't trust it. (Literally look at the parent article?)
I did have that time and capacity in the past. Maybe you didn't. But the landscape is, to me, objectively different now.
I fully agree that tech workers have never been less empowered. When I said that individuals have never been better equipped to pursue quality, I meant within projects they control. There will likely never be agency again at a large company.
We're on the same page about most of this, but my perspective differs in a couple of ways: - I have worked at companies that didn't prioritize quality and had unreasonable deadlines for most of my career. It's the default state of corporations with leverage over their workers. - This was getting worse before AI. Big tech was the anomaly in terms of pay, perks, benefits, etc, and they were already jointly deciding not to do that anymore around the time of Musk's Twitter layoffs. The main reason they had offered such a compelling environment was to prevent talented people from being hired by their competition, in which case they were perfectly happy to let them "play around" and focus on quality.
I find it frustrating when people imply that employment was an idyllic wonderland before AI ruined everything. AI is another step, potentially the final step, in boiling the middle class. It didn't invent poor working conditions, it brought tech workers down closer to everyone else's level.
The best move right now is to move to a small startup or start something on your own. I fully understand that this is easier said than done, when you have bills to pay and deadlines keep you too exhausted to escape. To some extent, that's probably part of why the deadlines exist.
Self respecting engineers care also in the age of AI.
Exactly. I cared, and the amazing engineers that taught me cared. We didn't have KPIs in 2008, we had code reviews and mean senior engineers (that were secretly kind) that would tear your code apart and send you back until you got it right.
And we were allowed to? I raised so many tickets over time that had no real business impact beyond "this is a security risk" and nearly all of them I was allowed to adress.
Maybe that makes me some kind of tech-boomer, like people responding to me are implying, but I don't really give a shit.
>I can't state my disdain enough for the rabid, careless advancement of AI, and all the bros that keep telling me to catch up or fall behind, while they are actively causing harm to the hundreds of thousands, or millions of users they have, without care or consideration.
Have you ever considered that the people who are pro-AI are not a monolith, and that it can be simultaneously true that you should absolutely be learning how to effectively and safely use LLMs in engineering workflows, while it is also true that there are people who are using them very unsafely?
Beig opposed to learning how to use AI correctly because some people use it recklessly is like being opposed to learning AWS in 2015 because some people left sensitive data in open S3 buckets.
Do not confuse negligent misuse of the tool for the value proposition of the tool itself.
Re-read my post before quoting it and getting defensive. I'm not anti-AI, I use Claude daily, partly because that's all the industry is now, partly because it can be an amazing tool.
I'm furious about slop-shovelling profit-driven development that actively causes harm to people. And the lack of care for, or outright disdain of customers, specifically fuelled by a misuse of AI, and the people telling me to get used to this, the people telling me to stop reading my own PRs, the people telling me this is just how it is now. When we all know that doesn't have to be true.
Just bide your time until there are some absolutely catastrophic breaches like what is happening in Japan right now on a daily basis, and then people and legislators will slowly start demanding accountability.
I just don't see any grassroots movement advocating for responsible software development winning out with a demographic that can't even unionize for their own self benefit.
There is no part of me that believes people will fundamentally do the right thing without an incentive to do so (money) or punishment (fines, jail). There won't be action until there is so much suffering that governments can no longer ignore it.
Yeah, makes sense.
Only counterpoint is we have very well functioning unions in Norway for engineers.
But yeah, I don't see grassroots movements winning anywhere. Not even here.
https://en.wikipedia.org/wiki/Norwegian_Society_of_Engineers...
https://en.wikipedia.org/wiki/Tekna_(Norway)
It's surprising how different the standards are between software engineering and all other kinds of engineering, for instance civil or mechanic.
The old joke was "If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization", but in the AI era it feels like a butterfly would be enough.
Counterpoint: if programmers wrote programs the way structural engineers designed buildings, ecommerce apps would take half a decade to build and the software industry would have just gotten around to inventing Windows Vista and jQuery in the last year or so.
Maybe that's okay.
If you're saying e-commerce would simply be consistent and secure product catalogs with fewer promotional and tracking dark patterns, web scripting would rely on a and extensible and functional library for manipulating markup documents, and a stable OS kernel with few dark patterns and no cloud accounts, many of us would be sold in a heartbeat.
I don't think you're making the counterpoint you mean to be.
Don't be naïve - the industry wouldn't be any less laden with surveillance crapitalism; the presence of that isn't a product of the pace of development, it is a product of the nature of consumers to eagerly suffer limitless abuse so long as the product "just werks", looks/feels polished, and doesn't require any kind of learning or discomfort.
Freedom from everything you enumerated already exists for anyone who cares enough to pursue it - robust ad and script blockers, operating systems with no coercion or user hostility, simple and stable scripting languages. They're not as polished as the products of manipulative and extractive abusers like Microsoft, Google, Apple, Discord, et al., and accordingly, users shun them, for the reason I explained above.
> it is a product of the nature of consumers to eagerly suffer limitless abuse
Ah, there's the victim blaming.
> Freedom from everything you enumerated already exists for anyone who cares enough to pursue it
Do you have any idea how daily life is for non-technical people? It isn't reasonable in our society to expect everyone to switch to FOSS, and blaming 60 year olds for getting their identity stolen because they weren't using a Linux phone or rolling their own self hosted matrix chat server is such a cruel and unbelievably out of touch take.
Hey, if the option is on the table for states to fight back, up and including enacting the corporate death penalty (forced dissolution) against Microsoft, Alphabet, Meta, Apple, etc, I'm all for that.
That would never happen though, the victims would plea and fight too vigorously in defense of their own abusers.
Who should be blamed for the actions of corporations ultimately subject to the society-scale approval of their consumers? No amount of misbehavior, malice, negligence, or abuse from the company can alienate the modal user of their products, they've thoroughly demonstrated this too many times to count.
It's not the doddering elderly folks I blame the most, it's everyone younger than them, the folks who grew up with the technology, aware of every abuse by these companies, who refuse to change their behavior. The ones who know better. The ones who can explain to you in great detail all about the surveillance, the dark patterns, the emotional manipulation, the extractive commercial angle, and who all refuse to adjust their behavior anyway. It's a choice, and it's one they make on a continual basis.
And to answer your question, no, I don't have much of an idea how daily life is for non-technical people besides them voluntarily suffering abuse like theft of their time (advertisements), or voluntarily paying a bribe to avoid the abuse, for no real reason other than that they don't care enough about the abuse or the bribe to spend 5 minutes learning how avoid both for free for the rest of their life.
> And to answer your question, no, I don't have much of an idea how daily life is for non-technical people
This is very clear. You're incredibly disconnected from how most people in society operate, their motives, and capacity.
And your malice and contempt is entirely misplaced.
You seem confused.
malice /măl′ĭs/ noun
1. A desire to harm others or to see others suffer; extreme ill will or spite.
2. The intent to commit an unlawful act without justification or excuse.
3. An improper motive for an action, such as desire to cause injury to another.
None of the above apply to me. I don't want to harm or injure anyone, and I don't want anyone to suffer. Consumers choose suffering and choose to voluntarily accept harm and injury, and I wish they wouldn't.
But you can't force people to not harm themselves, it's something they have to choose for themselves. Anyone who's ever lived with or loved an alcoholic or drug addict understands this. The contempt isn't for the person, it's for the theft of the person's faculties. The response to the person being unwilling to change is frustration with their choices, not contempt for the human being.
Hmmm but that software would also last a century so… your point? Moving fast and building unstable software isn’t really that great.
You won’t lose your job to AI...but to someone using AI...and they will lose everything...to someone who knows cybersecurity.
I think this more or less sums it up.
Interesting times.
Regarding losing jobs - what I hope will keep me employed: opiniated creativity and common sense. Not sure if I'm right.
if those are the reasons... I'd feel very uneasy
These past few years, it feels as though the brake pedal has been replaced with another gas pedal.