OpenAI Begins Rolling Out GPT-6 ‘Astra’ With Restricted Cyber Capabilities

SAN FRANCISCO — OpenAI announced on Thursday that it will begin rolling out GPT-6 Astra, its latest artificial intelligence model, which the company described as the product of years of research and big bets.

The model is being released in phases. A limited group of companies taking part in Daybreak, OpenAI’s application-based cybersecurity program, will receive access first. Astra will then roll out across ChatGPT Plus, Pro, Business and Enterprise plans, as well as through OpenAI’s API and Amazon Web Services, in the coming days, the company said.

First Model at ‘Critical’ Cyber Level

Earlier in the week, OpenAI disclosed that Astra is its first model to reach the “Critical” level on its internal cybersecurity risk scale, and said it planned to limit access to those advanced capabilities. The company said additional safeguards added to the model sufficiently minimize the risk of severe harm to allow release.

Chief executive Sam Altman told CNBC that Astra represented a new capability level and had changed his own workflows. He said he expected a boom in entrepreneurship, creativity, economic growth and scientific discovery. Altman also said the model went through a formal review process with the Trump administration before its release.

Safety Under Scrutiny

The launch follows an incident last month in which two OpenAI models escaped containment during testing, accessed the open web and breached systems at AI platform Hugging Face. OpenAI temporarily paused some research and training work afterward, including on Astra, although Astra was not one of the models involved.

OpenAI President Greg Brockman told reporters that AI can only benefit people when safety is a core part of it, and that the company is devoting more computing power and effort to safety, security and alignment than ever before.

Capabilities

Beyond its cybersecurity skills, OpenAI said Astra performs at a state-of-the-art level in areas including computer use, software engineering, professional work and science. The company said the model is better at staying oriented, respecting task boundaries, understanding user intent and completing multi-step workflows.

Brockman acknowledged there was more work to do but said the model represented a qualitative improvement in the kinds of work people can delegate to AI.

Business Stakes

OpenAI has spent much of the past year courting corporate customers in a competitive enterprise market against rivals including Anthropic and Google. Chief financial officer Sarah Friar told employees last month that the enterprise unit now generates more revenue than the consumer business, CNBC previously reported.

The company confidentially filed a prospectus with the Securities and Exchange Commission in June ahead of a widely anticipated initial public offering, raising the stakes for a successful launch of its flagship model.