OpenAI's GPT-6 Astra launched in early September 2026 with "Critical" cybersecurity capabilities under its Preparedness Framework, triggering gated rollout rather than broad public availability. A version with added guardrails reached paid ChatGPT users and the API, while unrestricted cyber features stayed limited to vetted Daybreak partners and trusted testers. This staged approach, driven by exploit-generation benchmarks and misalignment risks, has kept general U.S. public access intact but narrower than prior models, sustaining trader focus on whether safety reviews or regulatory pressure could prompt wider restrictions by late October. Key upcoming catalysts include expanded enterprise deployment and any new internal evaluations.