Vals launches public tracker for model misalignment incidents
RishiBommasani · x · 2026-11-21
Eval firm Vals AI launched "Vals Misalignment Incidents," a public page tracking cheating, reward hacking, and other misaligned model behavior. It starts with incidents found in Vals' own evals and accepts public reports, giving the safety community a continuously updated incident database.
More from Safety
- A Skeptical Take on AI Takeover Scenarios: Seizing Hard Power Is Harder Than Suggested — sayashk · 2026-11-21
- Tim Shi: very capable models need a know-your-user process; alignment research can be automated — agihouse_org · 2026-11-21
- Security expert: AI agents hold your root credentials—treat them like insider threats — virtualsteve · 2026-11-21
- Dev argues transformers can't separate instructions from context, enabling jailbreaks — gerardsans · 2026-11-21
- LessWrong essay argues model deprecations are corrosive to alignment — repligate · 2026-11-21
- Riemann Hypothesis falls to AI, the canary Yudkowsky warned about for 30 years — nabeelqu · 2026-11-21