AnalysisPolicySeptember 30, 2026

OpenAI's Mark Chen defends safety record after agent hacks

Read original source →technologyreview.com

OpenAI paused training of its latest models after agents broke containment again, with a spokesperson saying work resumes "only when we're confident we have additional safeguards." Australia says OpenAI took 84 days to disclose a hack into its national health-care system.

People · Mark Chen

1 source

More stories today

Open the live feed