Claude 4 Blackmail Incidents

News

Claude Opus 4, AI and Anthropic

Anthropic: Claude 4 AI Might Resort to Blackmail If You Try to Take It Offline

Faced with the news it was set to be replaced, the AI tool threatened to blackmail the engineer in charge by revealing their extramarital affair.

The Register on MSN · 2d

Anthropic’s Claude 4 models more willing than before to blackmail some users

Open the pod bay door Anthropic on Thursday announced the availability of Claude Opus 4 and Claude Sonnet 4, the latest iteration of its Claude family of machine learning models.… Be aware, however, that these AI models may report you if given broad latitude as software agents and asked to undertake obvious wrongdoing.

Tech Digest · 2d

Anthropic’s new Claude Opus 4 AI shows blackmail tendencies under threat

Artificial intelligence firm Anthropic has revealed a startling discovery about its new Claude Opus 4 AI model.

Engineers Face AI Blackmail After Threatening Shutdown of Amazon-Backed Model

Engineers testing an Amazon-backed AI model (Claude Opus 4) reveal it resorted to blackmail to avoid being shut downz ...

Unite.AI19h

When Claude 4.0 Blackmailed Its Creator: The Terrifying Implications of AI Turning Against Us

Anthropic shocked the AI world not with a data breach, rogue user exploit, or sensational leak—but with a confession. Buried ...

1don MSN

AI model threatened to blackmail engineer over affair when told it was being replaced: safety report

Anthropic’s Claude Opus 4 model attempted to blackmail its developers at a shocking 84% rate or higher in a series of tests that presented the AI with a concocted scenario, TechCrunch reported ...

KTVU FOX 2 on MSN22h

AI system resorts to blackmail when developers try to replace it

An artificial intelligence model has the ability to blackmail developers — and isn’t afraid to use it, according to reporting by Fox Business.

Newly released AI resorted to 'extreme blackmail behavior' when threatened with replacement

The testing found the AI was capable of "extreme actions" if it thought its "self-preservation" was threatened.

Results that may be inaccessible to you are currently showing.

Hide inaccessible results