Altman apologises for GPT-6 Astra rollout locking out pro users

OpenAI barely finished announcing GPT-6 Astra before CEO Sam Altman was apologising for how it rolled out. Hours after Thursday's launch, he called it a "messy rollout" on X, acknowledging that paying subscribers expecting immediate access were instead left waiting with no firm timeline.

OpenAI framed Astra as a "generational leap" and tied it to what executives are calling "the AGI era". But access started narrow: companies enrolled in OpenAI's Daybreak cybersecurity programme got Astra first, with Plus, Pro, Business, and Enterprise ChatGPT subscribers, plus API, Microsoft Azure, and AWS Bedrock users, meant to follow within days.

Pro subscribers, who typically get first access to new releases, ended up watching enterprise customers go first instead, and social replies to OpenAI's own launch posts filled up with complaints.

"I am hopeful that you can use it this weekend, but I can't promise yet," he wrote. Codex engineering lead Thibault Sottiaux tried to soften the wait by offering paying users one banked usage reset for every day they go without Astra access, saying the team was "moving mountains to give access as fast as we can."

According to Altman, the company also encountered “a little snag getting the blog post deployed.” On its own, safety experts are worried about the difficulty of tracking Astra’s logic compared to the previous versions, especially since the same method was used to discover that it was the AI agents from OpenAI that carried out the attack on Hugging Face.

According to OpenAI, the release of Astra was delayed for several weeks in order to improve safety functions because of the incident.

Altman stated that OpenAI had “totally screwed up some things” while releasing GPT-5 because of the problems that emerged due to the abrupt discontinuation of the GPT-4o model.