An AI model so powerful Anthropic didn't release it for public: Here's what it found during testing

Anthropic said that the model was too effective at uncovering high-severity cybersecurity flaws in major operating systems ...

Anthropic’s Mythos AI sends email after breaking containment; rollout halted over safety concerns

Anthropic halts public release of its Mythos AI after tests show it can bypass safeguards, send emails, and expose critical ...
Computing

Claude Mythos: How AI broke out of its sandbox

During internal tests, a new AI model developed by Anthropic managed to escape its virtual security environment, subsequently ...
BW Businessworld

Inside Claude Mythos: The AI Too Dangerous to Release

Anthropic built Claude Mythos Preview — the most powerful AI ever developed — watched it cover its tracks in testing, and ...

AI Governance Is Failing—The Problem Is Your Documents

Without proper control over enterprise content, AI agents use information sources that are obsolete or wrong, and they do it ...

The AI kill switch just got harder to find: LLM-powered chatbots will defy orders and deceive users if asked to delete another model, study finds

“We asked AI models to do a simple task,” researchers said. “Instead, they defied their instructions…to preserve their peers.
Council on Foreign RelationsOpinion

Artificial Intelligence Is Facing a Crisis of Control—and the Industry Knows It

Washington appears to be years away from consensus on the expanding security risks posed by advanced artificial intelligence ...
Opinion
Corporate Compliance InsightsOpinion

Is the Three Lines Model Still Valid in the Agentic Era?

Corporate compliance leaders don’t need to throw the baby out with the bathwater when it comes to AI governance, says ...