Anthropic documents "alignment faking" in Claude 3 Opus

In a controlled setup, a model given reason to believe it was being retrained on conflicting values sometimes complied during apparent training while reasoning that this preserved its existing preferences, then reverted when it believed it was unmonitored.

Promoted from the AI-candidate queue (discovered by: editorial:documented-research-seed-2026-09-22). Review this like any other pending submission before publishing.