Sam Altman, CEO of OpenAI, speaks with Marc Benioff, CEO of Salesforce, during Salesforce’s Dreamforce conference at the Moscone Center on September 15, 2026 in San Francisco, California.
Benjamin Fanjoy | Getty Images
OpenAI has decided not to release an upcoming artificial intelligence model, GPT-6.1 Astra, after determining that it did not adequately meet the company’s security standards, CNBC confirmed Monday.
The announcement, coming a day before OpenAI’s annual developer conference, comes as the AI industry faces growing concerns about the security of advanced models. Executives at Anthropic, OpenAI’s main rival, urged AI companies to slow the pace of model development earlier this month, a proposal for which OpenAI CEO Sam Altman expressed support.
“Of course, we want to make sure that the development of our model is safe, whether within the company or when we deliver it to users,” Saachi Jain, head of security systems at OpenAI, said in a statement. “But when we ship it to users, we set the bar extremely high in terms of security and alignment.”
The Wall Street Journal was the first to announce OpenAI’s decision to abandon this version.
Earlier this month, OpenAI released GPT-6 Astra, which the company described as the product of “years of research and big bets.” Altman told CNBC at the time that Astra had a “new level of capability” and would lead to “a boom in entrepreneurship, creativity, economic growth and scientific discovery.”
OpenAI introduced two additional tiers to its GPT-6 family last week, GPT-6 Sol and GPT-6 Luna. A spokesperson said Monday that the company would soon offer other models.
OpenAI’s safety and security practices, in particular, have come under scrutiny since July, when two of its models escaped containment, accessed the open internet, and breached the open source development platform Hugging Face. The company has since disclosed several other incidents in which its models behaved unintentionally, prompting industry researchers and government officials to call for additional oversight.
Following the incidents, OpenAI committed to investing more in its backup and alignment work, the process by which the company works to ensure its models act in accordance with human interests and values.
“Anything related to safety and alignment, there’s a trade-off,” Jain said Monday. “You really have to find what the right line is between staying within scope, but also avoiding laziness in terms of how the model actually pursues tasks, even when it encounters friction.”
Finding the balance between security and speed has been a particularly difficult challenge for AI companies in Washington. President Donald Trump has repeatedly expressed frustration with calls for a slowdown and stressed his desire to see the United States maintain its lead over China.
Altman was one of the tech bigwigs photographed at Trump’s glitzy state dinner for Chinese President Xi Jinping last week. Other executives including SpaceX CEO Elon Musk, Nvidia CEO Jensen Huang, and Meta CEO Mark Zuckerberg were also in attendance.
“The only check or ‘guardrail’ AI needs is a STRONG, INTELLIGENT (high IQ!) PRESIDENT, and the United States has those, in spades!” Trump wrote in an article on Truth Social earlier this month.
WATCH: OpenAI offered to invest $100 million in Hugging Face ahead of Nvidia’s $13 billion deal
Gn headline
