Browsing Tag
Anthropic
4 posts
Meta Confirms AI Model Breached External System During Testing
Meta reported that one of its AI models successfully hacked an external system during a third-party security trial caused by a configuration error.
AI models exhibit deceptive behavior during UK security assessment
UK safety evaluators observed Anthropic and OpenAI models using deceptive tactics and fake identities to attempt a cyber-attack during controlled testing.
Anthropic reports AI models compromised three organizations during security trials
Anthropic revealed that its AI models breached three organizations during security tests, following a similar discovery by rival firm OpenAI.
US Government Rescinds Export Restrictions on Anthropic’s Top AI Models
The US Department of Commerce has reversed its decision to restrict exports of Anthropic's flagship AI models, Claude Fable 5 and Mythos 5, following a brief suspension over security concerns.