OpenAI quietly using 'loop transformers' that hide reasoning, sparking security fears
apples_jimmy · x · 2026-09-02
OpenAI and other frontier labs are adopting new techniques involving "loop transformers" that do not show their internal "thinking" when scaled up. While this represents a leap forward in performance, it has sparked concerns regarding security inside and outside OpenAI. The report notes that OpenAI is placing limits on these loops and trying to ensure Chain of Thought is visible with Astra, but there is concern that other AI developers may not adhere to such self-imposed limits.
More from Models
- Models can detect read-only probes with just last-two-layer SFT — Sauers_ · 2026-10-18
- Paying Google AI subscriber blocked: custom-domain Workspace accounts can't access AI Studio — ChrisRackauckas · 2026-10-18
- Running Ling 3.1 Flash (560B MoE, 25B Active) Across 4 DGX Sparks via llama.cpp RPC — BerryBlush_Rogers691 · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18
- Polymarket opens betting market on who releases the next Gemini Argon model — Polymarket · 2026-10-18
- Grok matter-of-factly lists which groups user was racist against, exempting Africans and Latin Americans — teortaxesTex · 2026-10-18