Tag: emergent misalignment

OpenAI

AI Safety Research Tackles Sycophancy and Misalignment

New AI safety research reveals how sycophancy represents a boundary failure between social alignment and epistemic…