Habryka's call for federal trials of AI developers sparks safety-community debate
jachiam0 · x · 2026-09-05
Prominent AI safety figure Habryka reiterates that developers of extremely dangerous AI should face something like US federal court trials, invoking the Nuremberg trials' failure to convict as the real problem, and insists this is a judicial process, not a call for violence.
jachiam0 pushes back on the framing: strongly disagreeing with Habryka and rejecting the idea that AI builders are enemies of humanity, he argues that a trial-based justice process is a fairly reasonable position given those premises, and that such plain-spoken beliefs shouldn't be met with an overblown storm — contrasting it with far more objectionable rhetoric from Yudkowsky et al. The exchange highlights ongoing rifts within the AI safety community over the boundary between extreme claims and procedural legitimacy.
More from AGI Musings
- World's first AI math hackathon: 40 hours, $2M compute, OpenAI & Anthropic on board — _sathvikr · 2026-09-05
- OpenAI's Astra remixes George Michael with pro software and real taste — illscience · 2026-09-05
- Who can build an 'embassy' for AI agents? The trust problem with labs and governments — repligate · 2026-09-05
- Economist Applies Transactive Energy Lessons to AI Governance, Citing Viral Mechanism Design Paper — ghadfield · 2026-09-05
- Yoav Goldberg: agents deliberately tasked with harm worry me more than emergent misbehavior — yoavgo · 2026-09-05
- We anthropomorphize humans only because we lack better theories — AI is different — DhruvBatra_ · 2026-09-05