Musk proposes AI safety peer review: Anthropic tests OpenAI, everyone breaks everyone's models pre-release
elonmusk · x · 2026-09-15
At the All-In Summit, Elon Musk offered a practical framing of the AI safety problem: OpenAI and Anthropic are close enough in capability that neither can slow down without handing the lead to the other. He noted Anthropic puts more care into safety than OpenAI, yet even Anthropic insiders publicly worry about their own models.
His proposed fix instead of voluntary restraint: cross-testing. Anthropic runs its safety test harness on OpenAI models, OpenAI tests Anthropic, SpaceXAI tests both, and leading Chinese AI companies join the same system — everyone tries to break everyone else's models before release. The full discussion is in the latest All-In episode, alongside Starship, space data centers, and Terafab topics.
Related event: Musk Proposes Cross-Lab Pre-Release Safety Testing Among AI Companies(6 posts)→
More from AGI Musings
- Detecting covert instruction-following via internal representations parallels J-space work — dioscuri · 2026-11-02
- Paras Chopra: Mastering AI means building self-improving systems, not solving problems directly — paraschopra · 2026-11-02
- Databases to distributed systems: every systems skill is merging into AI inference and hardware — Abhishekcur · 2026-11-02
- Limitless analogy: smarter people may gain far more from superintelligence — McDonaghMatthew · 2026-11-02
- Hunting for the most concrete "curing cancer with AI" roadmap of 2025 — iskander · 2026-11-02
- Avoiding AI brain rot: Miles Cranmer's lecture on learning with AI — _ScottCondron · 2026-11-02