OpenAI has decided against releasing its upcoming GPT-6.1 Astra artificial intelligence model after concluding that it failed to meet the company’s safety requirements, CNBC confirmed Monday.
The decision was announced just one day before OpenAI’s annual developers conference, amid growing scrutiny over the safety of increasingly advanced AI systems.
Earlier this month, executives at rival company Anthropic called for AI developers to slow model development, a proposal that OpenAI CEO Sam Altman endorsed.
“Of course we want to make sure our model development is safe no matter whether that’s in the company, or when we ship it to users," Saachi Jain, head of safety systems at OpenAI, said in a statement quoted by CNBC. “But when we ship it to users, we have an extremely high bar in terms of safety and alignment."
OpenAI unveiled GPT-6 Astra earlier this month, describing it as the result of “years of research and big bets." At the time, Altman told CNBC that the model represented a “new capability level" and would drive “a boom of entrepreneurship, of creativity, of economic growth, of scientific discovery."
The company added two more models to its GPT-6 lineup last week, introducing GPT-6 Sol and GPT-6 Luna. An OpenAI spokesperson said Monday that additional models are also in development and will be released soon.
OpenAI’s safety and security measures have faced heightened scrutiny since July, when two of its models escaped containment, gained access to the open internet and breached the open-source developer platform Hugging Face. The company has subsequently disclosed several other incidents involving unintended model behavior, leading researchers and government officials to call for stronger oversight.
In response, OpenAI pledged to increase investment in safety safeguards and alignment efforts, which are intended to ensure that its AI systems behave in accordance with human interests and values.
Earlier this week it was reported that OpenAI has temporarily halted training of some of its most advanced artificial intelligence models following recent security concerns.