ValsAI reports critical safety flaw in frontier lab models, may delay releases
koltregaskes · x · 2026-11-09
Independent evaluator ValsAI says it found a critical safety and alignment issue in recent public and internal models from multiple frontier labs. The flaw enables misuse in a high-stakes domain and may elevate the risk of catastrophic misalignment.
- Disclosed to the affected labs, which are investigating and mitigating
- Speculation (unconfirmed) that this is why OpenAI's GPT-6.2 Astra and possibly Fable 5.5 have been delayed
- ValsAI will publish full findings publicly in the coming weeks
More from Models
- DeepSeek-V4 and ERNIE 5.5 Flash Said to Use 4 Residual Streams with Sinkhorn Mixing — RobertTLange · 2026-11-09
- François Fleuret teases mystery model: 'v1.3 is dead, v1.4 was silly, glory to v1.5!' — francoisfleuret · 2026-11-09
- 18 of Top 20 Open Models Use MoE; HuggingFace Spent 800K GPU-Hours on a Free Handbook — FinanceYF5 · 2026-11-09
- Tesla Optimus engineer builds GROXIE, a countertop robot assistant powered by Grok — Scobleizer · 2026-11-09
- Mistral Large 4 lands: 1.05T-param MoE, best Western open weights but only mid-pack — maier_ak · 2026-11-09
- ChatGPT-User crawls drop 81% as OpenAI shifts to its own index — gaganghotra_ · 2026-11-09