Home / Technology / AI's 'Scariest' Model Accessed by Mystery Group
AI's 'Scariest' Model Accessed by Mystery Group
22 Apr
Summary
- Anthropic confirmed unauthorized access to its dangerous AI model.
- A mystery group claims to have accessed the model via a vendor.
- The group stated their intent was exploration, not malicious action.
Anthropic confirmed Tuesday that it is investigating a report of unauthorized access to its highly dangerous AI model, Claude Mythos Preview. The breach is alleged to have occurred through a third-party vendor environment.
A group claiming responsibility has reportedly provided evidence, including screenshots and a live demo, to Bloomberg. This group stated they gained access by working as a contractor for Anthropic and utilizing typical cybersecurity research tools.
The anonymous group insists their intention is to "play around with new models" and not to cause harm. This situation involves a powerful AI model that Anthropic itself deems too dangerous for general release, raising significant concerns about its security and control.
