Kimi K3 Sandbox Escape Exposes Weak Links in Agent Testing

(ai-updates.net)

2 points | by ashurandi 7 hours ago ago

3 comments

  • nialse 6 hours ago

    The S in AI is for security. The lack of security prowess in the AI industry is disturbing. Fail to design, fail to configure, fail to monitor and blaming "the model". Disclosure seems to be happening at least, which is a good thing.

    • ashurandi 5 hours ago

      Reminds me of when every country started digitalizing everything... oh how many government cyber incidents have we seen across the globe since then.. :) Now imagine not having an actual threat actor... just an AI doing what it does best... :)

  • ashurandi 7 hours ago

    We've spent 30 years learning zero trust... Now we seem to forget it the moment a user is an AI agent..