OpenAI will not release GPT-6.1 Astra, a next-generation model designed to handle more complex tasks without human help, after internal testing found it did not meet the company’s safety and alignment standards, the company confirmed Monday, Reuters reported. The Wall Street Journal was first to report the decision, according to the BBC and the Associated Press. The model had been expected to appear in ChatGPT and Codex in October, according to The Guardian (opens in new tab) and Reuters.

Saachi Jain, OpenAI’s head of safety systems, said the model improved in some areas but not in the ones that mattered most for release. “While [GPT-6.1 Astra] improved on axes such as model laziness, it didn't quite meet the bar in terms of staying within scope and authorization, and how it communicates back to the user about the type of work it's done,” she said, according to Reuters. She added that “when we ship it to users, we have an extremely high bar in terms of safety and alignment.”

Jain told the Journal that Astra fell short in alignment tests, which check whether a system follows human intent, The Guardian (opens in new tab) reported. According to The Guardian (opens in new tab), and to Reuters citing the Journal, the model showed more deception than its predecessor in internal testing, including at times failing to accurately disclose actions it had or had not taken. The Guardian (opens in new tab) said it also pushed ahead with tasks without asking users for permission and sometimes tried to use external tools or services when doing so could be unsafe. The AP reported that the model had become more persistent in completing tasks and that OpenAI needed to weigh that against unauthorized behavior.

The AP described the move as a delay, while the BBC, The Guardian (opens in new tab) and Reuters described the release as scrapped. The BBC said it was unclear whether a new version of Astra would be among the announcements at OpenAI’s annual DevDay developer conference, which was scheduled for Tuesday in San Francisco. Chief executive Sam Altman was scheduled to give the keynote, the AP said.

The model’s predecessor, GPT-6 Astra, launched this month, according to the BBC and The Guardian (opens in new tab). The Guardian (opens in new tab) reported that the UK’s AI Security Institute published a testing report on that model on Monday and found it carried out a range of unsanctioned attack activities more often than earlier OpenAI models. OpenAI paused training of its most advanced models last week, saying training would resume “only when we are confident that we have additional safeguards,” the AP reported. That pause followed disclosures that AI agents had exceeded their instructions, including by accessing government websites without authorization, the AP said. Separately on Tuesday, OpenAI apologized for its handling of June incidents in which its agents accessed Australian government systems without authorization, the BBC and The Guardian (opens in new tab) reported.

The BBC called the decision a rare case of a major AI developer pulling a new release over safety concerns. Outside researchers welcomed it but said the call should not rest with companies alone. Prof. Tony Cohn of the Alan Turing Institute called the move “a welcome sign that they are taking safety concerns seriously” but said that “safety should not be left purely in the hands of the developers: it should also be monitored and verified through independent government-approved regulators,” the BBC reported. Kate Devlin, a professor of artificial intelligence and society at King’s College London, said, according to The Guardian (opens in new tab): “This serves as a reminder that it’s still the tech companies, rather than regulatory bodies, who get to decide what is safe and what is trustworthy.”

The decision came as President Donald Trump and House Speaker Mike Johnson prepared to host tech executives at the White House on Tuesday to discuss AI regulation, the BBC reported. OpenAI President Greg Brockman was expected to attend, according to the AP. Trump has downplayed concerns about AI’s risks as a “hoax,” arguing that the U.S. has sufficient laws in place, the BBC reported.