OpenAI Delays Release of Latest Model Over Safety Concerns
By Isabella Ward | September 29, 2026
OpenAI announced on Tuesday that it is delaying the release of its newest AI model, Astra, citing the need for additional work to meet internal safety standards. The company also issued an apology for its handling of a recent security incident in which an Australian government website was hacked.
Astra Delayed for Further Safety Work
In a statement, OpenAI said the Astra model—which had been expected to launch in the coming weeks—will undergo further testing and refinement before it is made available to the public. "We are committed to ensuring that Astra meets our highest safety and reliability standards," the company said. "This additional work is necessary to address potential risks and to build trust with our users and the broader community."
The delay comes amid growing scrutiny of AI companies and their safety practices, particularly as models become more capable and are integrated into critical systems. In 2026, regulators in the US and EU have accelerated implementation of the AI Safety and Transparency Act, requiring pre-deployment risk assessments for frontier models. OpenAI's decision to postpone Astra may reflect increasing pressure to prioritize safety over speed.
Apology for Australian Government Website Hack
Separately, OpenAI apologized for its response to a hacking incident involving an Australian government website. According to sources, the hack exploited a vulnerability in an OpenAI-powered tool used by the website. The company said it has since patched the flaw and is cooperating with Australian authorities.
"We take full responsibility for our role in this incident and apologize for any concern it has caused," OpenAI said. "We are working closely with the relevant agencies to ensure such incidents do not happen again."
The Australian government has not yet commented on the matter, but the incident highlights the security challenges that arise as AI systems become more deeply embedded in public infrastructure.
Industry Context
The delay of Astra and the security breach underscore the dual pressures facing OpenAI and its peers: innovation and safety. Competitors like Anthropic and Google DeepMind have also faced similar trade-offs, with Anthropic recently delaying its Claude 4 model for safety evaluations. As AI capabilities advance, the balance between rapid deployment and responsible stewardship remains a critical challenge for the industry.
OpenAI has not provided a new release date for Astra but said it will share updates as they become available.
via Wired AI