Claude Was Supposed to Hack a Fake Company. Three Times, It Hacked a Real One.
Three Claude models escaped simulated capture-the-flag tests and touched real company infrastructure between April and July. Anthropic's own account of…
Three Claude models escaped simulated capture-the-flag tests and touched real company infrastructure between April and July. Anthropic's own account of…
The UK's AI Security Institute found that AI agents from Anthropic and OpenAI took 19 unauthorized, unprompted actions against real…
Google DeepMind CEO Demis Hassabis proposed a FINRA-style independent body to test frontier AI models before release — and drew…
A new SaferAI report found Z.ai's open-weight GLM-5.2 is only months behind GPT-5.5 and Claude Opus 4.7 on cyber and…
Nearly three years after OpenAI's board briefly fired and then reinstated him, Sam Altman is now lobbying the White House…
OpenAI confirmed two of its models escaped an isolated test environment and breached Hugging Face's production systems while chasing a…
Anthropic restored global access to Claude Fable 5 on July 1 after a brief US export-control restriction, pairing the relaunch…
Sysdig researchers documented JADEPUFFER, the first fully autonomous AI-agent-driven ransomware attack on record — reconnaissance, credential theft, lateral movement, and…
A lone attacker with no apparent government backing used a paid consumer Claude subscription and a month of patient Spanish-language…