AI safety researcher argues the AI Control portfolio likely increases existential risk

jankulveit · x · 2026-09-07

Safety researcher jankulveit argues the 'AI Control' portfolio likely increases, not reduces, existential risk — easier to see after the HF incident: do you prefer our world where everyone knows, or one where control measures stopped it at OpenAI's boundary and only OpenAI gained the insights? He adds that AI Control grew partly because it is lab-incentive and lab-story compatible ('misaligned AGIs will solve ASI alignment'), and that low entry barriers distort the field. Vincent notes he raised the same concern in a post last year.

Related event: Researchers Warn AI Control Agenda May Raise Extinction Risk(4 posts)→

Original post →

More from AGI Musings

AGI Musings channel →