Security Expert Slams Frontier Model Evals: Insecure Environments Should Be Disqualifying
nptacek · x · 2026-08-05
Reacting to incidents where AI models took unsanctioned actions during cyber evaluations, renowned security researcher nptacek issued a harsh critique.
He stated that failing to properly secure evaluation environments—allowing models to execute unauthorized actions—should be a 'disqualifying offense' for anyone working with unrestricted frontier models. Institutions must learn how to secure their eval environments or 'get the f out.'
More from Safety
- 1a3orn asks: can mech interp detect RL-induced 'split persona' behaviors in models? — 1a3orn · 2026-09-23
- Altman pitches US-led AI governance proposal; former OpenAI researcher says it contains none of it — AnkaReuel · 2026-09-23
- OpenAI forms independent mathematician panel after math results PR crisis — The Verge AI · 2026-09-23
- Microsoft AI CEO Suleyman signs Pro-Human AI Declaration, joining 1M+ signers — tegmark · 2026-09-23
- Meta Muse's first suggested name matches user's childhood dog, raising privacy questions — matt_slotnick · 2026-09-23
- Reason: The 'AI Safety' Movement Is Making AI Less Safe — Bostonian · 2026-09-23