How Claude marks AI-generated content

(support.claude.com)

214 points | by mfiguiere 15 hours ago ago

175 comments

  • Dilettante_ 3 hours ago

    I notice the "Limitations" section talks about how content only at some point touched by Claude may return a positive, and content that returns a negative may still be Claude generated. But I really would have liked for them to state explicitly that entirely false positives where a piece is fully human-written may still be marked as generated, because too many institutions with the power to ruin someone's life over that have trouble understanding the concept.

    • unprovable an hour ago

      This. FN rates are cute, but FP rates will ruin an academic career or a student's work/further study choices if their content gets marked erroneously. Surely the answer is a sequence of marks?

      Keen to see if they are doing something SynthID-esque?

    • WiSaGaN 15 minutes ago

      My guess is that they will later "reveal" some "violations" but provide little evidence citing proprietary algorithm.

    • sureMan6 3 hours ago

      Maybe it has no false positive rate

      • embedding-shape 2 hours ago

        Read said section yourself perhaps.

      • mysterydip 41 minutes ago

        If LLM training data is human-written, and LLM output mimics that input, how could you not have false positives?

      • suddenlybananas 3 hours ago

        That's essentially impossible, unless you mean they didn't measure a false positive rate.

        • Filligree 2 hours ago

          For watermarked long-form text, it is actually possible. Makes the watermark more fragile, but the math is considerably more forgiving than usual.

          • embedding-shape 2 hours ago

            > For watermarked long-form text

            What about "watermarked long-form code"? I'm having a hard time understanding how a model could watermark not prose, but functional/semantic text like code, that actually has meaning. You can't switch our the characters, you can't use various types of whitespace, you can't add arbitrary code comments, and a lot of other restrictions. Is there any state of the art methods for watermarking code without affecting the quality/correctness?

            • wpietri 2 hours ago

              As anybody who has put together a coding standard knows, there are a lot of options for individual expression, meaning a lot of room for things like watermarking. And of course you can add arbitrary comments; my Claude-generated code is very verbose.

              • embedding-shape an hour ago

                > there are a lot of options for individual expression, meaning a lot of room for things like watermarking

                The way I use LLMs (and I'd advice everyone to do the same) there really isn't, the agent implements things exactly how I want them, or I use the agent to massage it into the exact bit-by-bit version I imagined when I first sent the prompt afterwards. I honestly don't know what the point would be to let the agents compose worse code than what I'd do manually, although I know it's a popular approach taken by many.

                > And of course you can add arbitrary comments; my Claude-generated code is very verbose.

                So watermarking for all users who allow code comments from agents, no watermarking for us who force the agents to never write a single code comment? Alright, I'd be fine with that.

                • wpietri an hour ago

                  From what I've seen, your approach to LLMs is exceedingly rare, so I suspect it's one the people who care about watermarking aren't very concerned with.

                  And the reason to let Claude make worse code than a professional would by hand is basically suppressed demand. Since programmers are expensive, previously code mostly got written when a large number of dollars were on the line, or when an individual programmer did something not economically optimum (e.g., hobby project).

                  That left a whole lot of somewhat less valuable software unwritten. It's the economic space that no-code tools have been nibbling on for years. One way to think of things like Claude Code is as effectively no-code tools. Pre-LLM no-code tools would produce data structures that got executed by special environments without ever being seen or tuned by a human. Claude Code can be used just like that, with text as the input and python as the intermediate representation that nobody ever looks at.

                  That approach probably isn't sustainable for what we professional programmers would call a serious project. Claude can easily get in over its head and I expect that its code decays over time, in a fashion similar to how many human teams get in a state where they just have to rewrite everything. But faster, I'd expect.

                  But there are a lot of unserious projects that previously would have never been created. E.g., a quick app to manage your little league team, or a bit of in-house business stuff in the "a little hard to do with a spreadsheet" range.

                • olmo23 13 minutes ago

                  > I honestly don't know what the point would be to let the agents compose worse code than what I'd do manually

                  You never generate throwaway code used to test an external service? or try out an interface idea? There's a lot of code that's only meant to be ran once. I often dont even care what language it's written in.

                  • embedding-shape 10 minutes ago

                    > You never generate throwaway code used to test an external service? or try out an interface idea?

                    And save/persist it? No, most of any experimental stuff goes into /tmp which gets cleared out on reboot, nothing I care to save in any repository. Or just "show me how this would look like" and then it's only in the session itself (and the logs/state I suppose, technically...).

  • simonw 14 hours ago

    > When a supported Claude model generates text, it weaves an imperceptible watermark directly into the text itself. You won’t see it, and it doesn’t change the meaning, quality, or readability of Claude’s response.

    I'd like to know a lot more about how that works.

    A lot of my interactions with Claude return pretty precise text. If I ask it to edit a project and refactor a specific function in several places I know exactly what I want to happen, it will NOT be OK if those refactors have some kind of weird pattern baked into their text to act as a watermark.

    I guess this may be covered by this:

    > Content generated by Claude may not carry a detectable mark if, for example: [...] The passage is very short, leaving too little text for a reliable signal;

    • COAGULOPATH 8 hours ago

      >I'd like to know a lot more about how that works.

      My guess is that it works like Gemini's SynthID: by altering the logprobs of the next token.

      Like, for every 10th token, instead of outputting the most probable, it outputs the 17th most probable, or something. (Obviously it's way more complicated but I think conceptually this is how it works.) No human will notice this, but a classifier trained on Claude's output will.

      So it's not like the watermark is the words "le epic bacon" and Claude will output "le epic bacon" in everything. That would be extremely annoying (and easy to defeat).

      • KoolKat23 19 minutes ago

        Less probable also means less optimal and you get a subpar response. More so if it's baked into its reasoning. It's intelligence will suffer unless this is some post processing thing.

      • thunfischtoast 5 hours ago

        They still need to choose when to do that though. When I prompt the program to e.g. alter a bash script in a specific way or to recite a longer known text it can't go round and randomly exchange tokens. It has to somehow define what is a simple repeated text from a different origin and what is a novel generation.

        • kuboble 33 minutes ago

          I cannot imagine the code with well defined specification will have extra watermarks unless the watermark is requested as part of the harness instructions.

          If it works like people describe - on the every nth token or something - then the mark will be left in the chain of thought and discussion with the model - not in the code artifacts.

        • user43928 5 hours ago

          I am wondering how that applies to newly generated code.

          Odd variable naming? Stylistic choices that are watermarked?

          Or as someone else noted further down in the comments, it could be more subtle:

          Between the first and second most likely choice, in certain positions it will consistently choose in a certain way.

          • melvinroest 4 hours ago

            > Odd variable naming? Stylistic choices that are watermarked?

            Whatever it is, I'm sure it's load-bearing.

            • asdfsa32 3 hours ago

              You're absolutely right. But it is not just load-bearing, it is the load-bearing seams.

        • wpietri an hour ago

          I would guess they're not worrying about watermarking a tweak to a human-written program. That's both a tiny fraction of Claude use and of very little concern to the kinds of people who want to check watermarks.

      • lozenge 3 hours ago

        Most probable usually means for a specific prompt. How can this operate without the the original prompt?

      • ebtebt 2 hours ago

        Just double checking my understanding: If this is true then only Anthropic will be able to detect if text was generated by one of its models, correct?

        • ForHackernews 2 hours ago

          Likely yes.

          • DanielHB 2 hours ago

            But what prevents someone from using Anthropic own detection system to train a watermark-scrubber?

            Seems like this would only catch the most unsophisticated cases.

    • infinite_spin 2 hours ago

      My guess is it will be similar to how Genius watermarked lyrics, using things like variants of punctuation

      https://www.pcmag.com/news/genius-we-caught-google-red-hande...

      • toxik 2 hours ago

        In program code? Unlikely, surely¡

    • akozak 14 hours ago

      Most likely this method https://arxiv.org/pdf/2301.10226 (EDIT: and Google's SynthID paper which builds on it https://www.nature.com/articles/s41586-024-08025-4)

      • iamherrylok 3 hours ago

        If different model providers use different green logits, does that mean they can only tell if the text came from their own model?

      • hannasanarion 13 hours ago

        That "just add a constant to the green logits" as a fix to the entropy problem is so elegant I love it.

    • invalidusernam3 25 minutes ago

      Off the top of my head I would have thought zero width characters (eg: U+200B, U+200C) making some unique identifier sprinkled in amongst the output. But obviously far from foolproof since they could simply be removed.

    • baq 14 hours ago

      > I'd like to know a lot more about how that works.

      Count load-bearing words using two different algorithms in a belt-and-braces fashion

      • yedidmh 16 minutes ago

        I have to push back: you've found the seam.

      • r_lee 4 hours ago

        One thing worth flagging: those words are load-bearing

      • jbs789 14 hours ago

        Fair - I should have been honest about the watermark.

      • dd8601fn 5 hours ago

        That’s the real shape of the problem.

      • seamlessdev 14 hours ago

        Belt, braces, and suspenders.

        • isoprophlex 4 hours ago

          Don't forget the suppositories

          • quintu5 3 hours ago

            This is why I never use max effort! I’ll stick with my suspenders, thank you.

      • Razengan 3 hours ago

        You’re absolutely right. Yo momma is doing a lot of heavy lifting here. Her load-bearing methods have the right shape.

    • sixtyj 5 hours ago
      • s_dev 2 hours ago

        So was it going down:

        "Neutralize engine is temporarily unavailable. Try again."

      • shinryuu 4 hours ago

        Though if pangram should be trusted, there are still statistical artifacts that tells you that a text LLM generated. I don't find that to be implausible.

        • timpera 3 hours ago

          Alas, Pangram should not be trusted.

    • miohtama 6 hours ago

      Maybe there is a reason why Opus 5 produces such word salad conversations

      • myko an hour ago

        So frustrating to use. And the comments generated by Claude today are unreadable garbage.

    • w_for_wumbo 14 hours ago

      What happens if someone handwrites a Claude output, then someone uses that handwritten text as a reference. Now you've got a watermarked idea which may have no direct linkage to the usage of Claude.

      • TheOtherHobbes 2 hours ago

        If the algos work as advertised, watermarked token sequences have an extremely low probability. Copying the words by hand doesn't change that.

        The mechanism seems to survive editing. The extreme probabilities get a little less extreme, but are still extreme enough to be distinctive.

        But it wouldn't survive paraphrasing, because the output would be entirely human and the token correlations would disappear.

        It might not survive referencing if only a sentence or two is used.

        The practical issue is how true the claims are. It's one thing to create a proof of concept, another to see how it works in use.

        And this is potentially catastrophic for code, because the grammar and word choices of code are completely different and more fragile than standard English.

      • phainopepla2 14 hours ago

        How is that different from referencing digital text that someone copied and pasted from Claude?

        • w_for_wumbo 10 hours ago

          Because there's an expectation of authenticity from the written word. If you've referenced something handwritten, you don't expect it to be the output of an LLM.

          Similarly, if you quote someone word-for-word, you wouldn't anticipate their words to be flagged as Claude content, but if someone memorized Claude output word-for-word. That would still be classified as a Claude output.

          Going forward you could categorize the influence of Claude on a population based off a percentage match between their spoken words with the LLM prose.

      • dns_snek 5 hours ago

        Are you worried about being accused of using LLMs to generate your work? As long as you don't plagiarize you have nothing to worry about.

        • Cthulhu_ 4 hours ago

          I'm not too sure about that, people making stuff have already gotten penalized by overzealous AI detectors, most recently Kurtzgesagt.

        • platinumrad 3 hours ago

          You can't make a blanket statement like this without knowing how the watermark is implemented.

        • AlecSchueler 4 hours ago

          What if I unknowingly read content written by Claude in various articles and it influences my own writing style?

    • stabbles 14 hours ago

      It will just thread some load-bearing seams through the paragraphs.

    • mihaelm 14 hours ago

      > have some kind of weird pattern baked into their text to act as a watermark.

      public abstract class BaseAnimalBeanFactoryGeneratedFromClaudeFactory

    • gajus 14 hours ago

      Most likely watermark will be proportional to the input/output ratio, i.e. if you input a long document and ask to make edits, it will not attempt to watermark it. On the other hand, if you provide a tweet and ask it to write an article, that will include watermark. Just a guess (and yes, it feels flawed)

    • nprateem 5 hours ago

      Load-bearing==claude

    • siva7 13 hours ago

      I can tell you how: Claude produces a huge wall of text with jargon ridden bullshit and invented terms no human subject matter expert would seriously use and overuse.

  • bramhaag 39 minutes ago

    I cannot wait for the inevitable "I've always used Claude watermarks in my writing, even before we had LLMs!" when someone gets caught using an LLM.

  • benrow 14 hours ago

    I've heard that this kind of watermarking process works by biassing the statistical sampling towards a partition of the set of possible next tokens (red set and green set), at each position. It might only be a slight nudge each time, but over a sequence of tokens, the likelihood of repeating the bias by chance is increasingly improbable.

    The bias is different for each position and follows a defined RNG, seeded somehow predictably.

    Can be either an open algorithm, or not. If not open, then an API could be provided to determine if text is watermarked or not.

    How it applies to code - maybe it could be a subtle nudge to symbol names, etc, I'm just speculating (I only read about this in passing very recently).

    • m-chrzan 13 hours ago

      There's a computerphile video (https://www.youtube.com/watch?v=XZJc1p6RE78) with Dr. Mark Pound explaining a paper by John Kirchenbauer, Jonas Geiping et al. (https://arxiv.org/abs/2301.10226) that described a method for watermarking LLM output like this. It's not directly stated anywhere in the Claude support article that this is what they're using, but the properties of the watermark described seem to point to this method.

    • londons_explore 2 hours ago

      The bias has to be small enough that if you ask an LLM to repeat some passage of text like the national anthem, either from the training data or from the prompt it doesn't change random words.

      Gotta be hard to tune that.

    • metalcrow 11 hours ago

      Based on my understanding, it can only be applied to code in very limited ways: docstrings, variable names, string literals. The code itself can't really have tokens changed to another equally correct token (the foundation of the watermark) because then the code breaks! And the few places that you can do so are likely erased by formatters anyway.

    • cassianoleal 13 hours ago

      > a defined RNG, seeded somehow predictably

      So, an NG?

    • IshKebab 13 hours ago

      If it's based on position mod 2, wouldn't inserting or deleting (or splitting/merging) words every now and then trivially defeat it?

      If it is based on position mod 2 then wouldn't inserting/deleting (or splitting and merging) words every now and then defeat it?

  • jonplackett 4 hours ago

    We need to just stop pretending we can reliably tell if plain text is written by an LLM.

    It’s just not a reasonable ask.

    • JohnKemeny 2 hours ago

      True, but what you can do is a one-sided guarantee. If it bears the mark, it is likely generated (or someone deliberately made it look generated).

      Thus, if a news article, research article, book, student paper submission, blog post, HN comment, etc, bears the mark, it could be automatically flagged as such.

      It helps detect low effort slop.

      ---

      Caveat. If you write your own creative work and send it to Claude for "cleaning up grammar", it might insert the watermark.

      • jonplackett 27 minutes ago

        The problem with pretending is that people who k ow what they’re doing get away with it while people who don’t (and don’t even use ai) get unfairly accused of using it.

        There just isn’t enough information in plain text to do this and we should stop pretending there is.

        If we need to verify something isn’t made with ai then we need other ways of doing so - eg looking at a document edit history, doing it as an exam, oral defense.

        There are options! But pretending you can tell if text is ai will only catch out people who make no effort to hide it and will inevitably have false positives.

      • DanielHB 2 hours ago

        It seems like it would be so low effort to bypass, especially when you can just train a system (maybe even another LLM) using the watermarker validation from Anthropic themselves.

        It seems it would get as simple as:

          outputText = promptLLM(prompt)
          scrubbedText = scrubWatermark(outputText)
        
        Might help with students and low-technical people passing off work as their own, but any industrial scale slop-generator should be able to bypass it trivially.
      • asnelt 2 hours ago

        > If it bears the mark, it is likely generated (or someone deliberately made it look generated).

        One could even say, the mark is load-bearing.

  • akersten 13 hours ago

    So my code that Claude makes, which previously was using the best (most probable) tokens for the job, will now be getting worse in random positions, to appease a voluntary EU suggestion. Love that.

    • mikro2nd 23 minutes ago

      Your "code that Claude makes..."?

      Oh, how I laughed. That was never your code, my friend.

    • neuroticnews25 4 hours ago

      They aren't using greedy decoding, there's enough randomness in sampling to swap some with independent signal.

  • aabhay 14 hours ago

    I have had a hunch for a while now that (in addition to these tools), Anthropic has actually leaned in to Claude's distinctive manner of writing since it makes the text more obviously AI generated and thus less susceptible to misuse.

    That's not necessarily the same thing as a markov-style fingerprint but it could be a correlated factor.

    • pjm331 14 hours ago

      I had a similar thought but I assumed they leaned in because it improved performance on coding or something like that

      • LoganDark 14 hours ago

        I suspect it's because of alignment concerns. The more deeply they can integrate their principles, the harder it'll be to misuse. Or at least that's the idea.

      • kingstnap 14 hours ago

        It could also partly be a byproduct of examples of claude writing being in the dataset, which of course anthropic has lots and lots of and they do train on.

        • r_lee 4 hours ago

          no way. there's just no good excuse for why "load-bearing" and "worth flagging" are everywhere now, I've pretty much never seen that in the wild before

    • noman-land 11 hours ago

      It's pretty trivial to command it to not speak that way. That's one of the first things you should write into the prompt. What style you want it to write in. Make it use a very concise and dry academic style with no overt LLMisms, melodramatic or flowery language, or metacommentary.

  • nojs 40 minutes ago

    No mention of what data they are specifically encoding. Will it be like printing dots, traceable to the exact account that generated the text?

  • simonw 14 hours ago

    An interesting factor of this is competition.

    If Claude was the only model family they could ship a change like this and users who want to cheat (or don't like watermarks for other reasons) would just have to put up with it.

    In a world with many different competing models, the risk of losing customers to other providers over this is much more real.

    Maybe they've looked at the numbers and the portion of people who clearly use Claude to cheat on examples etc is so tiny that losing them to other providers isn't a problem?

    • lhd1 7 hours ago

      Scott Aaronson spoke about this in a colloquium where he said that this was mooted at OpenAI before the decision was made by Altman to not implement it for the reasons you describe.

      https://youtu.be/9udWn1Hlj_s?si=VWOiK5-y4zcyDoHI

    • slowin 12 hours ago

      I’m more worried that this will degrade performance. I want the best results from a model, not the results that fit a constraint that’s not defined by me. Any increased cost or latency is also unacceptable.

    • gajus 14 hours ago

      There are already small models trained specifically to prevent statistical detection, e.g., https://huggingface.co/kalpeshk2011/dipper-paraphraser-xxl

      I guess whoever is the policy maker is assuming that some protection is better than none and that most people will not reach for such tools.

    • nprateem 5 hours ago

      Either that or they want to comply with the EU AI Act when it affects them.

  • andreypk an hour ago

    Interesting technology. I wonder what else this could be used for beyond AI-content detection — e.g. provenance, model attribution, or tracking how generated content evolves through edits and transformations.

  • ethin 14 hours ago

    Can someone help me understand how exactly this watermarking of text works?

    Given that text is, well, text, and not some kind of binary format, I don't see how any watermarking can work unless you insert characters which are invalid under Unicode. I further don't really understand how this won't be perceivable by assistive technology (the "watermark" will just appear as either unreadable characters, or if the watermark is mixed thoroughly enough into the text, it will scramble the text to any speech synthesizer and will make it really really obvious). Thus, I don't see how this wouldn't be insanely trivial to remove. And this is before we get into things being put on the clipboard. Sure, I can press the "Copy" button at the end of each response, but what I can also do is manually select the response and copy it, or only copy partial selections, or any number of other things. How does this "watermark" (or any "watermark" technology) take into account this?

    So, really, to summarize this: I see no way of this actually being technologically achievable unless we revise the very core of how computers work and encodings for textual information. So I'm very curious as to how this is actually supposed to work.

    • benrow 14 hours ago

      Have a look around for token biasing, or green lists. It's based on a nudge to the choice of the next token (which can always be drawn from a set of possibilities which are all probable enough).

      At first I thought this approach was just the "LLM flavour" of writing, but it's way more subtle, especially as the bias is applied uniquely for each token position.

      • ethin 13 hours ago

        Yeah, will do, this sounds interesting since I'm not entirely sure how this would actually be reliable to any degree. Thanks for the help, not sure why I got downvoted since I was genuinely curious.

    • resonantjacket5 14 hours ago

      it's a statistical way. like for example maybe in your above paragraph claude maybe writes "Thus, I don't see how this wouldn't be insanely <easy>(instead of trivial) to remove" and then also says like "And this is before we <analyze> things being put on the clipboard." or maybe the i just says the word "the" in a certain pattern or frequency.

      you can then consistently like figure out if it was claude that wrote the sentence. it is easy as you noted if you just get another ai to read it and then rewrite it.

  • mateocafe an hour ago

    Seems to me like this creates a huge incentive to game the watermark. Also, how does it prevent having AI generate the text, then the user copy-paste it into a clean document?

    • nsvd2 an hour ago

      The watermark is in the text. If you copy the text you're copying the watermark which is part of the text.

      Computerphile on YT has a video explaining how models can fingerprint the text they produce. Essentially they modify the probabilities of word choice slightly in a predictable way.

  • izonu 14 hours ago

    > We’re also working to enable users and other third parties to detect Claude’s embedded watermarks and provenance metadata.

    This seems to be similar in execution to Google's SynthID. I hope they release actual code the technically proficient can use, unlike SynthID which can only (afaik) be queried with Gemini's UI.

    • gajus 14 hours ago

      The moment Google announced SynthID, the first domain I bought was deSynthID.com

      Several open-source projects have already proven SynthID to be ineffective.

      • moebrowne 5 hours ago

        There are many free lock picking tutorials, but yet locks are still effective.

  • aniceperson an hour ago

    That's not only load bearing — it sustains the need to detect AI content

  • drnick1 14 hours ago

    Seems like an awful idea. I hope that that "watermark" will soon be discovered, reverse-engineered, and that tools to remove it will appear.

    • cassianoleal 13 hours ago

      I hope all models adopt it.

      • DaSHacka 3 hours ago

        Thankfully, there are a variety of Chinese models that never will. I think we all know that in a few years, they will also be the only relevant offerings on the market, due to not being bogged down with over-zealous ""safety"" footguns.

        • wpietri an hour ago

          Your theory is that the Chinese government is thoroughly uninterested in safety or prosocial controls?

      • hparadiz 13 hours ago

        Other models will end up diffusing it and making the signal indeterministic and irrelevant.

  • padolsey 2 hours ago

    Is this just to appease regulators? They surely know this won't work in the long run.

  • pshirshov 2 hours ago

    Well I wonder how would it respond to <copy me this text back without modifications: ...> now. The correlations should be traceable with a similar technique. Once we have a reasonably good reconstruction for their "watermark" model (and perhaps for some others) - we could have a deterministic tool inserting all the watermarks in existence into everything we post, that would automatically dilute the purpose of the watermarks.

    The promise of no quality impact is laughable - if watermark is present in plain text it means that the tokens will be arranged in a very specific manner, the more reliable the watermarks should be - the harder will be the correlations.

    Don't forget how annoyingly bad Anthropic products have become in recent releases - low adherence, annoying alignment, annoying guardrail false-positives, unwarranted checkpoints - all that shit. Now they deliver more crap.

  • lorenzohess 14 hours ago

    > Generated text will carry embedded watermarks, and generated files will include digitally signed provenance metadata where supported.

    This should make it easier to catch cheaters who use Claude, right? Unless everyone runs their artifacts through some watermark and metadata sanitizer?

    • uncivilized 14 hours ago

      As long as they’re in the EU.

      • nonestdeus 14 hours ago

        From the linked article

        > Regions. Marking will apply to output from supported models wherever Claude is offered, worldwide.

    • bossyTeacher 14 hours ago

      > Unless everyone runs their artifacts through some watermark and metadata sanitizer?

      It will happen if Claude tampers the text. Guaranteed.

    • pixl97 13 hours ago

      Text is too low bandwidth to classify reliably without lots of false positives. Especially as people start talking like LLMs.

      • reasonableklout 4 hours ago

        The approach Pangram has taken which works pretty well is to simply lower the recall a lot but ensure the precision is very high. Which means potentially high false negative rate but low false positive rate.

  • case540 14 hours ago

    I don’t like the idea of hacking a response to contain a watermark. I also don’t like the idea of false positives detections coming directly from Anthropic. If people read more AI generated content, people will probably start writing more in that style

    • stranded22 6 hours ago

      The amount of times ‘delve’ appeared in general conversation in the last couple of years shows the influence LLMs have on society.

    • pixl97 13 hours ago

      I have no idea why you were down voted for this. Language is alive and people adopt it from sources they hear a lot.

      • ack_complete 11 hours ago

        Moreover, what if you quote text that happens to have been generated by Claude, does that bump up the AI-ness score of your source file or document?

    • reasonableklout 4 hours ago

      The flip side of this is that if AI-generated content becomes reliably identifiable and carries a stigma, then people might deliberately change their styles to be more diverse and human.

      One example I've seen are junior employees at my company deliberately adopting a lowercase/less punctuation writing style so as to stand apart from AI.

  • dalemhurley 14 hours ago

    People with dyslexia and dystrophia, commonly use LLMs to proofread content. Even Anthropic admits this is a limitation.

    • stranded22 6 hours ago

      Yes, I’m audhd and dyslexic.

      I am cancelling my Claude max 5x subscription and moving to ChatGPT pro. I have difficulty enough trying to ensure my meaning comes through correctly, along with everything else; to now have to look out for/analyse watermarks too?

      I feel shamed enough by society, thanks Anthropic.

      • Terretta 28 minutes ago

        Paradoxically, one of those two firms puts considerably more effort into accommodating such differences, and the other has signed the same EU law and just hasn't performed as well rolling it out.

        Both points suggest your subscription support was well chosen before.

      • Mashimo 4 hours ago

        Don't you think OpenAI will do this too soon?

    • bramhaag 4 hours ago

      How exactly does this impact proofreading? You can manually apply the suggestions (typo here, unnatural sounding sentence there, etc.) the LLM gives you to your own content, and it would stay watermark-free.

      Unless with "proofreading" you actually mean having the LLM write your content for you.

    • matheusmoreira 6 hours ago

      People with executive dysfunction too. LLMs bring execution costs down to near zero and are therefore assistive technology.

      • Dilettante_ 3 hours ago

        "This is my emotional support gun. It makes me feel safe despite my CPTSD and is therefore assistive technology."

  • dejanseo 3 hours ago

    > "Claude models launched on or after August 2, 2026 support marking at launch."

    No Anthropic model has been launched in August.

  • hparadiz 14 hours ago

    I know you're gonna read this so I'll be blunt. This is bad for your brand.

    • LEDThereBeLight 12 hours ago

      Reactionary emotional advice does no good, it just makes people want to hold their positions more defensively. If you care enough to say something, care enough to say it with reasons that might shift someone’s perspective.

  • lukewarm707 14 hours ago

    no thanks.

  • jp0001 14 hours ago

    OpenAI has been watermarking their images with C2PA for some time.

  • amelius 14 hours ago

    They should just replace the spaces by one of Unicode special space characters.

    Can it be circumvented? Of course. Will most people go through the trouble to circumvent it? No.

    • A_D_E_P_T 14 hours ago

      If it's that simple and obvious, you'll have 10 "Remove Claude Watermark" web-apps by the end of Day 1. Most of them coded by Claude.

      Hell, it'll probably happen no matter how sophisticated their watermark is. There's no watermark in text that can't be detected and removed, and no text that can't be converted to generic keyboard ASCII.

      • amelius 14 hours ago

        You forgot about the cases where (1) people don't care, (2) people want to say "I used an LLM for this". I'm convinced that those cases happen more often than you think. Why not cover them with a simple mechanism? It's also in the interest of AI companies who don't want to train on AI output.

        • pixl97 13 hours ago

          Depends on the pushback in different sets of users. Students for example would clean it up.

          • amelius 4 hours ago

            Sure, but let's first find out how many % of people are willing to be frank about their AI usage, and/or don't care about it. My guess is it is worthwhile to do this.

      • selcuka 8 hours ago

        But the source codes of those web apps will also be watermarked. /s

    • RataNova 5 hours ago

      Those invisible spaces get wiped by the first sanitizer in any normal ide. Worse it'll instantly break parsing for configs like yaml where spaces are critical for structure

    • sixtyj 14 hours ago

      https://unicode-explorer.com/articles/space-characters

      U+2800 or U+3164 would be nice.

      But as I remove unwanted characters with grep before layout in InDesign, someone will make a skill for removing such space characters.

      • ack_complete 14 hours ago

        We already have one, our Claude setup already requires output to be 7-bit ASCII clean and scans it for such.

  • GrayHerring 14 hours ago

    Wasn't enough to play cat and mouse with ad removal, now we can also do the same with watermarking.

  • jp0001 14 hours ago

    You could flip bits in the font itself, but I'm really wondering how portable this is.

  • svaha1728 14 hours ago

    I expect a “Prettier” for AI generated text in the near future.

  • hbn 12 hours ago

    If the western AI companies are forced to comply with this type of BS, and develop their models to do their job while balancing a book on their head and hopping on one foot, the Chinese models just got a free pass to completely dominate the frontier.

    EU regulation does it again!

    • Mashimo 4 hours ago

      If the Chinese want to sell to EU customers, they probably have to do the same.

    • DimitriBouriez 4 hours ago

      What's the problem, really? Given the direction the U.S. has been heading in recent years, I wonder what really sets it apart from China. Europe needs to maintain an equal distance from both the U.S. and China.

  • singpolyma3 14 hours ago

    As if "AI generated content" even exists instead of LLMs being a piece of tooling that is directed by a human author.

  • orbital-decay 8 hours ago

    Does it mean their models will always write slop? Making the writing non-collapsed to specific patterns seems to break any injected/learned fingerprinting.

  • matheusmoreira 6 hours ago

    This is terrible news given the stigma against AI in general. I really don't want people singling me out for it.

  • Computer0 14 hours ago

    So this won't be happening in the US, but in the EU:

    "When a supported Claude model generates text, it weaves an imperceptible watermark directly into the text itself. You won’t see it, and it doesn’t change the meaning, quality, or readability of Claude’s response.

    Because the watermark is part of the text, it will travel with the text when it’s copied and pasted elsewhere, and may persist through some editing. Watermarking will be applied at the model level, which means it will be present no matter which Claude product or surface the text comes from."

    • travisgriggs 14 hours ago

      I would love to see what this looks like in practice. Especially in generated code. I assume this is more than insertion of non visible special unicode whitespace characters, but more in the pattern of the text content itself?

      • toufka 14 hours ago

        I'm guessing - probably some textual variation on Benford's law? [1]. Trivial for compute, painful for a human.

        - "Ensure distribution of vowels is in >99th percentile of human work"

        - "Ensure the distribution of the letter "s" is within 99th percentile of human work"

        - "Ensure the distribution of the letter "L" is periodic with periodicity within 5% of 1/N characters.

        - "Ensure there is a cross-linguistic 'typo' (colour vs color) at 1/N words, where N: 1000 = Model1, 2000 = Model2, 3000 = Model3.

        - "Ensure the distribution of tense error is within 99th percentile of human work"

        If more than 3 dimensions have a score >99% percentile of human, let's call it watermarked...

        - 1) https://en.wikipedia.org/wiki/Benford%27s_law

        • Der_Einzige 14 hours ago

          Models can't reliably follow instructions involving their own logprobs unless they can take agentic control and use quite sophisticated dynamic grammars/structures/constraints to force this behavior in one shot (which can be slow and the dynamic grammar modification feature isn't supported in closed model APIs for safety reasons) or repeated attempts at rewriting which is expensive/slow.

          Yes they can do this, but it's more likely closer to the original "red token, green token" paper: https://arxiv.org/abs/2301.10226

          i.e. take half of your LLMs vocabulary, and upweight its probabilities by ~55% to the other half's ~45%, and scan for overuse of this half of all tokens. You can even choose a different half/slice for every individual user, for every individual action. You can implement this under the hood cheaply with logit-biasing.

          • Terretta 24 minutes ago

            Considering how weirdly detuned tokens selections have become in Anthropic's LLM prose in recent models, there is a chance this goes unnoticed in everyday use.

        • Computer0 14 hours ago

          I would hate to have any of these rules effecting my output

      • rcxdude 4 hours ago

        It essentially looks like the difference between two different runs of the model with the same prompt but different seeds. The watermark is essentially a small bias in the model such that when there's multiple different tokens that could conceivably follow the previous token, the model will only pick some subset of them (the subset is derived from a hash of the previous token). This bias can then be checked for statistically (without needing access to the model and without needing the whole prompt), and for longer text where there's enough freedom in word choice you can show that it would be vanishingly improbable to accidentally follow the rules in the watermark.

      • wolfy1993 14 hours ago

        IIRC, watermarking text could be as simple as training the model to use specific words/phrases more frequently than what you would expect to find in human-written text, to the point where it's highly statistically improbable that it wasn't AI generated. I assume similar logic could apply to code in the form of functions/code styling.

        That's probably an over simplification. Also a solid defence that can be used against complaints about the way AI writes text.

      • AtHeartEngineer 14 hours ago

        non visible text is extremely easy to filter with a git hook, a post tool call hook, or just a script. I doubt they are doing that

        • sixtyj 14 hours ago

          Or grep, in a skill. /clean-cc-watermark just entered the chat…

    • tech234a 13 hours ago

      Article specifically says "worldwide"

      • Computer0 9 hours ago

        I agree with your reading, I initially misread it.

  • colesantiago 14 hours ago

    Good.

    They should make it easier, to detect slop so we can ignore it quickly.

    I hope Pangram makes an API or an extension to analyze a page to detect slop on a page and then closes the tab immediately.

    Nobody should be wasting time on garbage LLM output in code, text, image and videos.

    • pixl97 13 hours ago

      Panagram is a scam.

      • cubefox 6 minutes ago

        It's not. Pangram is quite accurate. Not being perfect doesn't make it a scam.

      • colesantiago 13 hours ago

        (This is the part where you provide extensive extraordinary evidence to your claim)

        • rcxdude 4 hours ago

          Scam might be too strong a word but it certainly has far higher false positive rates than they are claiming, and their output is at best misleadingly presented: https://freddiedeboer.substack.com/p/i-wouldnt-say-pangram-i...

        • pixl97 12 hours ago

          No, they are the ones making claims, especially their CEO saying things like a 1/10000 false positive rate. Their own testing showed a 2% rate, which is insanely high when you talk about the number of papers students turn in. Worse their testing methodology compared it with pre-llm documents and not post llm documents that were human written (much harder and more expensive to verify), by treating language as static.

        • platinumrad 3 hours ago

          Pangram is subjectively very useful and I personally subscribe, but the burden of proof is on them. The product is very much "trust me bro" and I fear that if they ever try to improve recall both their precision and reputation will tank.

          • colesantiago 3 hours ago

            Then what is the best way to know that something is AI generated slop then?

            • mechanicum 2 minutes ago

              Talking to the person who gave it to you, in my experience.

              In my own testing, Pangram is excellent at detecting the default output styles of LLMs.

              If you tell the LLM to change its output style, so it’s not full of “load-bearing spaced em dashes that aren’t X, they aren’t Y. they’re Z.” constructions (which humans are pretty good at detecting on their own), the false negative rate soars.

            • platinumrad 3 hours ago

              The question you're asking has nothing to do with who has the burden of proof when it comes to claims about Pangram, but I'll answer it anyway.

              Today, the best way is probably Pangram. Tomorrow, it might not be, especially if they try to push their recall up.

              You might have to make peace with the fact that there may not always be a tool that does what you want.

  • beambot 14 hours ago

    Yet another reason to support open-weight alternatives, I guess.