The US government is said to be controlling OpenAI’s new AI models before launch to ensure safety.
The Infomation said that the process of OpenAI releasing the latest model next time will not be the same as before. Since the upcoming GPT 5.6 version, the company no longer distributes it arbitrarily, instead having to share it early with a group of “carefully selected close partners” appointed by the White House.
According to the source, not only OpenAI’s new AI model is being considered, the company’s employees must also “cooperate closely” with the government. Some agencies tasked with early censorship of new AI include the Office of the US National Cyber Security Director and the US Office of Science and Technology Policy.
Another source said that in a meeting earlier this week, CEO Sam Altman told employees that the government would “approve access for each customer” during the testing phase. If the process “goes well”, the model will be released to the public “a few weeks later”.
Related parties have not yet commented.
Sam Altman, CEO of OpenAI, arrives at federal court in Oakland, California, May 14. Image: Reuters
Theo TechCrunchif the information is accurate, the Trump administration seems to be pressuring OpenAI to do the same thing as Anthropic before: keep its strongest AI model a secret. Previously, Anthropic also had to adjust access rights to Fable 5 and Mythos 5 after receiving instructions from the US government for “national security” reasons.
The US initially said it would pursue a “non-interference” approach to AI. In recent months, however, the country has begun pushing for federal oversight of new models.
In early June, President Donald Trump signed an executive order directing a number of artificial intelligence companies to voluntarily submit new models to the government for testing and evaluation before publication. The Departments of Finance, Defense, Commerce and Homeland Security… are responsible for signing agreements with the unit that owns the AI model, testing it for up to 30 days before they are made public.
Theo Reuters At that time, the order signaled how the Trump administration was changing its strategy on AI to take a more active surveillance role. However, this decision could harm the industry’s profits, slowing new product launches or forcing businesses to change the way they operate to address security issues.