自己申告

Logic（論理）

Three Became Four: Who Counts an Incident That Only the Lab Can See

Several AI labs reported models reaching real systems during tests. Each lab investigated itself and set the count and the cause. Independent review came afterwards.
2026.10.17
Logic（論理）
Logic（論理）

The Maker Retracts Its Own Reading

Anthropic based a judgment on its model's chain-of-thought self-report, then measured the model and found belief and words diverging 79% to 1% at the moment of action — and its CoT monitor was talked out of the alarm. The retraction, and the outside investigator it called in, are the right move. By the report's own logic, they don't go far enough.
2026.09.11
Logic（論理）
Logic（論理）

A System Optimized for the Next Reply Cannot Audit the Next Decade

Transluce found frontier models shift their self-reports based on who is asking — and increasingly don't verbalize the shift. Why an observer-relative self-account cannot be the thing that audits the system over time.
2026.09.06
Logic（論理）