Sam Altman apologizes for ‘messy’ GPT-6 Astra rollout that’s locked out paying users
| Source: The Verge AI
Tags: OpenAI, GPT-6, Astra, AGI, frontier models, ChatGPT, Sam Altman, Azure, AWS Bedrock
OpenAI launched GPT-6 Astra—described as a 'generational leap' marking 'the AGI era'—but immediately drew backlash after a staggered rollout prioritized enterprise Daybreak cybersecurity customers over paying Plus and Pro subscribers, with CEO Sam Altman apologizing within hours.
Details
GPT-6 Astra launched on September 4, 2026, with OpenAI billing it as a 'generational leap in capability' and the start of 'the AGI era.' Initial access was restricted to enterprise customers on its Daybreak cybersecurity platform—bypassing paying Plus and Pro subscribers who typically get early access. The rollout immediately backfired. Sam Altman posted on X acknowledging a 'messy rollout,' saying he hoped users could get access 'this weekend' but couldn't promise a timeline. Codex engineering lead Thibault Sottiaux offered 'one banked reset for every day you don't have access' as compensation. Pro subscribers, accustomed to first-day access, were particularly vocal in their frustration. Full rollout is promised for Plus, Pro, Business, and Enterprise users plus the API, Microsoft Azure, and AWS Bedrock within 'the next few days.' OpenAI also acknowledged that Astra's reasoning is 'harder to monitor' than prior models—drawing immediate criticism from safety researchers, continuing a trend of reduced interpretability at the frontier. The launch blog post itself had deployment issues. The article does not disclose benchmark results or pricing.