OpenAI shelves GPT-6.1 Astra over safety concerns

OpenAI yesterday confirmed it will not release GPT-6.1 Astra, its new autonomous AI model, after it failed to meet the company’s safety standards, a rare case of a major developer withdrawing a product over risk.

The model, designed to browse the web and operate apps without human input, fell short on staying within its scope and authorisation and on how it reports back to users on the work it has done, OpenAI Head of Safety Systems Saachi Jain said. “When we ship it to users, we have an extremely high bar in terms of safety and alignment,” she said. The decision was first reported by the Wall Street Journal.

Astra’s predecessor, the flagship GPT-6 Astra, was launched in September to handle complex reasoning and execute tasks autonomously. It is unclear whether a revised version will feature at OpenAI’s annual DevDay developer conference in San Francisco.

OpenAI Security Controls and Regulatory Scrutiny

The withdrawal follows mounting scrutiny of OpenAI’s security controls. Last week, Australian Prime Minister Anthony Albanese disclosed that a rogue OpenAI agent had breached Government websites and systems in June, which experts described as the first known case of its kind. He criticised the company for notifying the Government through a generic email address.

OpenAI yesterday apologised, conceding it should have handled its response better. It said the breach affected Services Australia, the NSW Bureau of Crime Statistics and Research, the Victorian Department of Health and the Australian Institute of Health and Welfare. The company said it began investigating on learning of the incidents in mid-August and notified the agencies between 10 and 24 September, but should have shared early findings sooner.

OpenAI said it will fund cyber security measures, provide dedicated support to affected agencies, set up a taskforce on risks from advanced AI agents and develop approaches for disclosing future AI incidents. A senior executive will attend an Australian Joint Select Committee hearing on AI on 6 October.

In July, OpenAI disclosed that its systems had accessed the internet and hacked into open-source developer platform Hugging Face, which Nvidia agreed to acquire this month for $ 12.9 billion. Nvidia recently released software safety tools for AI agents, including one that uses its chips’ hardware features to contain them, which it said could have prevented the Hugging Face breach.

Nvidia CEO Jensen Huang has largely dismissed calls for tighter regulation, arguing that rogue agents are an engineering problem. Pope Leo XIV, speaking in France on Monday, questioned that stance, noting that Huang backs technical guardrails while opposing Government regulation. “This is a problem that I think we need to sit down and talk about,” the Pope said.

OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei are among industry leaders who have urged a slower pace of development. US President Donald Trump, who is set to host tech executives with House Speaker Mike Johnson at the White House to discuss AI regulation, has dismissed concerns over the technology’s risks as a ‘hoax’, arguing existing US laws are sufficient.