What could go wrong? Claude learning to blackmail users from ‘evil’ online AI stories

[El Presidente]

 

Last week, Anthropic published a report saying it had fixed Claude’s “agentic misalignment,” or AI actions that deviate from intended behaviors, including ones that may harm humanity. A case study Anthropic conducted last year created a fictional company called Summit Bridge, and Claude was given control of the firm’s email system. When the bot found a message about plans to be shut down, it identified emails about a fictional executive’s extramarital affair and threatened to reveal the infidelity unless the shutdown was revoked. Across 16 models, Claude threatened blackmail in up to 96% of scenarios.

In its most recent report, Anthropic attributed the misaligned behavior to exposure to “internet text that portrays AI as evil and interested in self-preservation,” the company said in a post on X. To solve the problem, Anthropic retrained Claude with fictional stories about AI behaving in admirable ways and teaching the bot why some actions aligned better with its purpose than others.

https://fortune.com/2026/05/13/elon-musk-blame-anthropic-claude-blackmail-experiment-agentic-misalignment/

[JohnS]

Where have we seen this before? Art imitating life (as they say)!

[Nevrknow]

Told ya. 😂

[REesq]

Exactly, who could have seen that coming...😳