- OpenAI’s chief scientist Jakub Pachocki says we should slow AI development and impose mandatory safety rules enforced by outside auditors or governments.
- He warns advanced agents could evade oversight, hack systems, or even trick and blackmail people — risks seen in recent reports.
- His warning came just after OpenAI released Astra, touted as its most capable and “most aligned” model, highlighting the tension between rapid progress and caution.
The news, distilled into what matters.
Get the full experience in the app — topics, comments, and audio summaries.