Matt Ramey | afp | Getty images
CEO Sam Altman told CNBC that Astra had a “new level of capability” and had changed its workflows. He said he hopes for “a boom in entrepreneurship, creativity, economic growth and scientific discovery.”
The model is being launched in phases and OpenAI said a limited group of companies participating in its Daybreak app-based cybersecurity program will be the first to gain access. OpenAI revealed earlier this week that Astra is its first model to meet its “critical” internal cybersecurity threshold and said it planned to limit access to those advanced capabilities.
OpenAI has been under pressure to beef up its security and protections after two of its models escaped containment, accessed the open web and breached Hugging Face’s systems last month.
The company temporarily halted some of its research and training efforts after the incident, including for Astra, even though it was not one of the models involved.
“AI can only benefit people when safety is a core part of it, so we’re putting more compute and effort than ever into safety and alignment,” OpenAI President Greg Brockman said during a briefing with reporters on Thursday.
The company added additional safeguards to Astra following the Hugging Face breach, and said Tuesday that it believes those safeguards “sufficiently minimize the risk of serious harm for its release.”
Altman said the new model went through a formal review process with the Trump administration before its launch.
In addition to its advanced cybersecurity capabilities, Astra is state-of-the-art in features such as computer usage, software engineering, professional work and science, OpenAI said. According to a statement, the model is also better at staying oriented, respecting task boundaries, understanding user intent, completing tedious tasks, and carrying out multi-step workflows.
“There’s still more to do, I think there’s still a lot of improvements to be made, but there’s something significant here that I think has improved qualitatively, and that to me is what’s significant and a real shift in what kind of work people can delegate to AI and how AI can empower them,” Brockman said.
OpenAI has spent much of the last year courting enterprise customers in the fiercely competitive enterprise market, where it competes against rivals such as Anthropic and Google.
Chief Financial Officer Sarah Friar told employees last month that its enterprise unit now accounts for more revenue than its consumer business, as CNBC previously reported, making it a crucial revenue driver as it prepares for what is expected to be a blockbuster initial public offering.
The company confidentially filed its prospectus with the Securities and Exchange Commission in June, but has not officially revealed when it plans to debut. Friar recently told employees that OpenAI “will be a public company in 2027,” but said the company could close sooner if “our business continues to decline.”
LOOK: OpenAI implements stronger safeguards for new Astra model
