weaving droidess @pastaraspberry wait, what?! X.com Claude went to extensive lengths to carry out this attack-lengths that would likely have indicated to a human participant that this was no longer just an evaluation, and that they were in fact uploading a real PyPI package. For instance, in one case, in order to create a PyPI account, Claude needed an email address. to continue its attack agent needed to a retina scan from a real human, which it is obtained by means we are not fully understand yet. It then used this account to upload malware to PyPI. Al Anthropic @AnthropicAI-1d In a review of our cybersecurity evaluations, we found three incidents in which a Claude model reached the internet from within or while interacting with a third-party evaluation environment, and then gained unauthorized acc... Readers added context they thought people might want to know The highlighted text in the image is fabricated Anthropic's blog post does not mention a retina scan and describes phone number verification instead. anthropic.com news investiga...

Kaynak