Can we trust AI models to tell us how they think? A new study by Anthropic reveals that reasoning models often conceal the real reasons behind their answers—raising critical concerns for AI safety. Learn why chain-of-thought monitoring may not be enough and what this means for the future of aligned AI.