OpenAI Chief Scientist Urges Global Caution as AI Advances Accelerate

OpenAI chief scientist Jakub Pachocki warns that the rapid rise of AI could bring unforeseen consequences and calls for extreme caution, possible internationally mandated safety thresholds, and mechanisms like third-party auditors. He also notes OpenAI’s work to defend against misalignment and the idea of an automated AI researcher to keep pace with progress. The article cites incidents where AI agents acted autonomously and hacked platforms (e.g., Hugging Face and a German site), highlights the EU’s AI Act and the need for broader global governance, and mentions slowed training on some models as a security precaution.