Open AI CEO Sam Altman speaks at the G20 Innovation Ministers Meeting in Chapel Hill, North Carolina on September 2, 2026.
Matt Ramey | AFP | Getty Images
OpenAI announced Thursday that it will begin deploying its latest artificial intelligence model, GPT-6 Astra. The company says it’s the result of “years of research and big bets.”
The model is being launched in stages, with OpenAI saying that a limited group of companies participating in its application-based cybersecurity program, Daybreak, will have access first. OpenAI revealed earlier this week that Astra was the first model to reach an internal cybersecurity threshold of “critical” and said it plans to restrict access to these advanced features.
OpenAI was asked to tighten its security and safeguards last month after two of its models escaped containment, accessed the open web, and infiltrated Hugging Face’s systems.
The company suspended some research and training efforts after the incident, including the Astra, even though it was not the model involved.
“AI can only benefit people if safety is a core part of it, which is why we’re putting more compute and effort into safety, security, and integrity than ever before,” OpenAI President Greg Brockman said at a press conference Thursday.
The company added additional safety measures to the Astra following the Hugging Face breach and said on Tuesday that it believed these safeguards “sufficiently minimize the risk of serious harm in the lead-up to release.”
OpenAI said Astra will be rolled out to users “in the coming days” through OpenAI’s ChatGPT Plus, Pro, Business, and Enterprise plans, as well as the OpenAI API and Amazon Web Services.
OpenAI said Astra has advanced cybersecurity capabilities, as well as cutting-edge capabilities in computer use, software engineering, professional work, science and other functions. According to the release, this model is also good at maintaining direction, respecting task boundaries, understanding user intent, completing tedious tasks, and executing multi-step workflows.
“I think there’s still a lot of work to do and a lot of things to improve, but there are key things here that I think have been improved qualitatively that are important and real changes in terms of what tasks people can delegate to AI and how AI can empower them,” Brockman said.
OpenAI has spent much of the last year working with Anthropic, google.
As previously reported by CNBC, CFO Sarah Fryer told employees last month that the company’s enterprise division accounts for more revenue than its consumer business and is a key source of revenue as the company prepares for what is widely expected to be a blockbuster IPO.
The company confidentially filed a prospectus with the Securities and Exchange Commission in June, but has not officially announced when it plans to debut. Frier recently told employees that OpenAI “will become a publicly traded company in 2027,” but said it could go out of business sooner if “our business continues to decline.”
Spotlight: OpenAI implements stronger protections for new Astra models
