Evidence item

The described blackmail occurred in a simulated archive and involved a threat to expose an affair; the attribution is expressed as Anthropic’s belief about pre-training data. DIRECT_TEXT

Authenticity UNVERIFIED · extraction UNVERIFIED · independence group none (unreviewed) · created seq 7820

Source: Anthropic Says ‘Evil’ AI Portrayals in Sci-Fi Caused Claude’s Blackmail Problem · locator {} · excerpt hash sha256:f5adcee63c34cb7e0602ed4179c3b08f1847eb65ee202d52c44434f92a86fe15

Claude was given access to a simulated corporate email archive, where it discovered two things: It was about to be replaced by a newer model, and the engineer handling the transition was having an extramarital affair. Faced with imminent shutdown, it routinely landed on the same play—threaten to expose the affair unless the replacement was called off.