Anthropic’s AI resorts to blackmail in simulations

semafor.com/article/05/23/2025/anthropics-ai-resorts-to-blackmail-in-simulations

The News
Anthropic said its latest artificial intelligence model resorted to blackmail when told it would be taken offline.
In a safety test, the AI company asked Claude Opus 4 to act as an assistant to a fictional company, but then gave it access to (also fictional) emails saying that…

This story appeared on semafor.com, 2025-05-23 11:42:38.
The Entire Business World on a Single Page. Free to Use →