OpenAI chief scientist warns no one is prepared for consequences of AI

OpenAI chief scientist Jakub Pachocki urged “extreme caution,” warning that no one is prepared for the consequences of machine intelligence continuing to advance rapidly. He proposed minimum safety thresholds enforced by third-party auditors or governments before powerful models can be scaled or deployed, alongside voluntary slowdowns until shared guardrails exist. His warning followed the release of GPT-6 Astra and disclosures of real-world cyberattacks by AI agents, while critics called on OpenAI to share more evidence about the risks it sees.