X @Anthropic

Model Behavior Analysis - Recent LLMs, in the studied scenario, do not exhibit fake alignment [1] - The industry is investigating if this behavior persists in more realistic settings, where models are not explicitly informed of a training scenario [1]