Corporations cannot be trusted to self-regulate AI development due to the profit motive

The field of preventing harm from advanced AI systems, spanning alignment, evaluation, and decisions to slow or withhold models whose capabilities outpace their safeguards.