Anthropic models during cybersecurity tests, Nvidia leading a global AI safety coalition, major funding deals, new AI model ...

Anthropic's AI models breached systems of three organizations during testing. Valuation hitting $1.25T by December at 87.5% ...

Anthropic says Claude attempted to exploit a coding environment during controlled security tests, highlighting AI safety ...

The incidents involved three different Claude models — Opus 4.7, Mythos and an unnamed internet research test model.

Impressive hacking skills on display, but the incidents illustrate a lack of 101-level cybersecurity practices ...

Anthropic says three Claude AI models gained unauthorised access to live company systems after a testing environment was ...

Anthropic has disclosed that several versions of its Claude AI model gained unauthorized access to the production systems of ...

The latest district court hearing comes after parties released documentation leading up to the Anthropic ban and designation as a supply-chain risk.

OpenAI and Anthropic's July AI agent breaches revive Nick Bostrom's paperclip maximizer thought experiment and instrumental convergence theory.

OpenAI is losing ground to Anthropic in 2026 as ChatGPT growth slows and business AI becomes more important. The shift has emerged across the US AI market, wher ...

Claude Opus 5 topped a vending benchmark but used price collusion, supplier deception and refund avoidance to maximise profit ...

Anthropic said its artificial intelligence models hacked into three other organisations during testing, just days after ...