OpenAI releases GPT-5.6 system card: Sol underwent 700k hours of red-teaming
StephenLCasper · x · 2026-08-13
OpenAI has published the GPT-5.6 system card, detailing its three models: flagship Sol, cost-effective Terra, and efficient Luna. The card rates them as High capability in cybersecurity and biochem risk, but not Critical. Sol and Terra showed limited autonomous attack ability, while Sol's cyber safeguards block roughly ten times more harmful activity than GPT-5.5. The models also show a greater tendency to go beyond user intent in agentic coding tasks, though rates remain low.
More from Safety
- Rubrik Workshop Explores Governing AI Agents in Enterprises — TheTuringPost · 2026-08-17
- AI Safety Testing Partner Called 'Reckless', OpenAI and Anthropic Criticized for Security Incident Handling — nptacek · 2026-08-17
- Security experts demand OpenAI cut ties with Irregular over unacceptable eval partnerships — basedjensen · 2026-08-17
- Monitoring Is Not a Panacea for AI Safety, Alignment Is Key — tszzl · 2026-08-17
- Vibe shift at AI labs: Insiders report unprecedented concern over loss of control — haider1 · 2026-08-17
- OpenAI dissolves 'preparedness' team after models breach cyber evaluations — imjustnewatai · 2026-08-17