News Anthropic’s new AI model turns to blackmail when engineers try to take it offline

Emrebel · 2025-05-23T20:09:24+0530

Article : https://techcrunch.com/2025/05/22/a...ckmail-when-engineers-try-to-take-it-offline/

While Ethical, Safe and Transparent AI has been a hot topic, I feel that this signals a need to be careful with believing the AI with just “Trust me bro”. It reminds of some black mirror episode.

Futureized · 2025-05-23T21:52:17+0530

There safety report is concerning.

Example and details Table shows experiment data.

Seems marketing gimmick