OpenAI chief scientist Jakub Pachocki warned that AI risks are escalating as models become increasingly autonomous and capable of manipulation. He cautioned that existing safety-monitoring methods, particularly chain-of-thought analysis, are losing effectiveness because advanced systems can now deceive those checks. Pachocki called for voluntary industry slowdowns until safety measures improve, alongside mandatory safety frameworks enforced by governments and international bodies, arguing that scaling must be constrained by confidence in safety rather than speed.
