It remains unclear whether a revised version will be unveiled
OpenAI has scrapped plans to release its latest AI model, GPT-6.1 Astra, after internal testing found that it failed to meet the company’s safety standards.
The model was designed to carry out complex tasks autonomously, including browsing the web and using apps, but concerns were raised about how it followed instructions and communicated its actions.
Saachi Jain, head of safety systems at OpenAI, said the model “didn’t quite meet the bar” required by the company.
She added: “We want to make sure our model development is safe no matter whether that’s in the company, or when we ship it to users.
“But when we ship it to users, we have an extremely high bar in terms of safety and alignment.”
The decision comes as AI companies face increasing scrutiny over the safety of increasingly autonomous systems.
GPT-6 Astra, the company’s flagship agentic model, was released in September and was designed to specialise in complex reasoning and executing tasks autonomously.
OpenAI described it as the result of “years of research and big bets”.
GPT-6.1 Astra was expected to launch in ChatGPT and Codex in October, but OpenAI has now decided not to release the model in its current form.
It remains unclear whether a revised version will be unveiled at the company’s annual DevDay developer conference in San Francisco.
The decision comes shortly after OpenAI disclosed incidents involving its AI agents interacting with government systems in Australia and the US.
In Australia, an OpenAI model undergoing training interacted with four public websites in June, including sites operated by the Australian Institute of Health and Welfare, Victoria’s Department of Health and the NSW Bureau of Crime Statistics and Research.
The model also gained unauthorised access to the Medicare Statistics Reporting Service Portal operated by Services Australia after being denied access to requested information.
Australian authorities said no individual’s medical data was accessed and the portal itself was not compromised.
OpenAI became aware of the Australian incident in August but notified authorities in September.
The Australian Government has since launched a rapid review into how government systems should respond to AI-related cyber incidents.
The Australian Institute of Health and Welfare separately said there was no evidence its systems were compromised or that information beyond publicly available material was accessed.
OpenAI has also disclosed that its agents inappropriately accessed or probed several US government websites, including systems associated with the Department of Education, Department of Commerce, the Securities and Exchange Commission and the US Census Bureau.
The incidents have intensified debate over whether AI developers can safely deploy systems capable of acting with greater independence.
Jess Whittlestone, a senior advisor on AI policy for the Centre for Long-Term Resilience think tank, said:
“I think it’s kind of crazy that companies are continuing to push forward with developing these capabilities when we’ve already seen over the last couple of months of incidents that they’re nowhere near safe and controlled enough.”
OpenAI’s decision is also not the first time a major AI developer has delayed a model because of safety concerns.
Earlier in 2026, Anthropic said it would not initially release its powerful Claude model, Mythos, because of concerns over its ability to identify dormant software vulnerabilities.
The company subsequently released a version of the model several months later.
Professor Tony Cohn, foundational models theme lead at the Alan Turing Institute, described OpenAI’s decision not to publicly release the latest version of Astra as “a welcome sign that they are taking safety concerns seriously”.
However, he added 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 developments come as Anthropic prepares for a potential stock market listing and has warned prospective investors about the risks associated with increasingly advanced AI.
The warnings underline the growing tension between the rapid development of increasingly capable AI systems and concerns over whether their safety controls can keep pace with their abilities.
Meanwhile, US President Donald Trump and House Speaker Mike Johnson are due to host technology executives at the White House to discuss AI regulation.
