Researcher Claims Technical Safety for Closed AI Models is Solved, Blames Incidents on Corporate Shortcuts
StephenLCasper · x · 2026-08-13
AI safety researcher Stephen Casper argues that technical safety for closed-weight AI systems is, at this point, a solved problem. He defines technical safety as the challenge of writing a robust safety specification and getting the model to align its behaviors with it.
He points out that SOTA safety tools are already capable of stopping most technical violations. However, the real-world failures of closed-weight model safety stem from the fact that AI companies take too many shortcuts during deployment. Consequently, he states that he no longer works on safeguards for closed-weight models.
More from AGI Musings
- 'Will you end humanity?' AI's candid answer becomes a clown meme — dermflork · 2026-10-16
- Stanford professor: AI-done homework scores high, students fail in-person finals — FinanceYF5 · 2026-10-16
- Data brokers pay $100K–$2M for SaaS datasets that may train their own replacements — FinanceYF5 · 2026-10-16
- tszzl: Pretraining science is like steam engines before thermodynamics, slouching toward interpretability — tszzl · 2026-10-16
- eigenrobot: AI acceleration will collapse human capital investment — eigenrobot · 2026-10-16
- As AI decompilation and vibe coding surge, who trusts AI-cloned software? — Honest_Risk3903 · 2026-10-16