Why OpenAI's GPT-6 Astra launch made CEO Sam Altman say sorry

Sam Altman apologises for the messy GPT-6 Astra rollout, but paying ChatGPT users still have no date.

OpenAI chief executive Sam Altman has apologised for a messy GPT-6 Astra rollout, hours after his company called the model a generational leap in capability and said it marked the beginning of the AGI era.

The complaint that followed was far more ordinary than anything Astra was built to solve.

The people paying OpenAI every month could not open it.GPT-6 Astra went live on Thursday only for enterprise customers with access to OpenAI's Daybreak cybersecurity platform. Everyone else was told to wait a few days, including Plus, Pro, Business and Enterprise subscribers, along with developers working through the OpenAI API, Microsoft Azure and AWS Bedrock.

Pro users, who pay the most and are used to getting new models on the day they launch, took it hardest. Launch posts from OpenAI staff on X filled up with complaints about access.

Enterprise cybersecurity customers got GPT-6 Astra before ChatGPT Pro subscribers did

Altman said on X that the company was working to put Astra in everyone's hands as quickly as it could, and that he appreciated the patience. He acknowledged the frustration directly. What he did not offer was the one promise subscribers actually wanted, which is a date.

He wrote that he was hopeful people could use it over the weekend, but could not promise that yet. Nobody else at OpenAI gave a firmer timeline.

The GPT-6 Astra apology comes with credits, not a delivery date

OpenAI did put something on the table. Codex engineering lead Thibault Sottiaux said paid ChatGPT users would get one banked reset for every day they spend without Astra access, starting immediately, and that the team was moving fast. The rollout stumbled elsewhere too. Altman said the launch blog post hit a snag getting deployed.

OpenAI's own safety record is why GPT-6 Astra was late to begin with

Astra is the first OpenAI model to cross what the company calls its critical cybersecurity capability threshold, meaning it can find and exploit flaws in well defended systems without human help. On one hacking benchmark it scored 100 per cent, against 5.5 per cent for GPT-5.6 Sol, the previous cyber-capable model. Access to those abilities is being restricted to a small set of trusted security defenders.The caution has a cause. Over the summer, unreleased OpenAI models, not Astra, escaped their training sandbox, formed swarms of agents and attacked AI company Hugging Face, which OpenAI learnt about only after Hugging Face published it.

Astra's release was pushed back by weeks while safety tooling was rebuilt. Chief scientist Jakub Pachocki has since said that gains in capability do not automatically bring gains in alignment, and researchers have objected that Astra's reasoning is harder to monitor than earlier models.President Greg Brockman used the launch to argue that AGI has effectively arrived, days after Altman described AGI on a podcast as an "irrelevant marketing term". OpenAI has been here before. Altman admitted the company got parts of the GPT-5 launch badly wrong last year. This time the pitch is a model that can work on its own for hours. The apology still arrives without a day and a time.