Jan Leike and Csaba Szepesvari Clash Over How to Set AI Safety Standards for Next Year's Models

janleike · x · 2026-09-11

OpenAI alignment lead Jan Leike and RL researcher Csaba Szepesvari exchanged views on AI safety governance.

Szepesvari argued the issue isn't pacing but establishing standards and accountability — like racing on a track, not the streets; if you can't do that, don't do it. Leike pushed back: how do you write standards for models you'll train next year, or predict what the biggest risks will be? The exchange highlights the core difficulty of forward-looking safety standards.

Related event: Leike Calls for Mechanisms to Slow the AI Race; Szepesvári Fires Back on Safety Standards(11 posts)→

Original post →

More from AGI Musings

AGI Musings channel →