Deception in AI

Anthropic Paper Probes Introspection in LLMs, Finds Limited, Inconsistent Signals

Researchers stress this is not evidence of self‑awareness, with reported effects fragile.