Recently, the US government has begun to strengthen controls over advanced artificial intelligence (AI) models under the pretext of "national security". According to a report by the US news website Axios on June 25, the White House has instructed OpenAI to release its next-generation model GPT-5.6 in phases. Before the new model is widely released, it can only be made available to a limited number of government-approved partners.
Sources revealed that the U.S. government believes the performance of OpenAI’s latest model is comparable to Anthropic’s Mythos model. Due to concerns about security, the White House National Cyber Director’s Office and the Office of Technology Policy have asked OpenAI to limit the scope of its new model’s release. This is the first time the U.S. government has requested AI companies to limit the release of new models.
According to sources, these advanced AI models are highly powerful. Therefore, the White House wants to ensure that the companies involved take adequate security measures. “This is the problem faced by such models at present.”
On June 17 local time, during the G7 Summit, U.S. President Trump and OpenAI CEO Orterman attended a work meeting.
OpenAI CEO Orlteman has agreed to restrict the release of GPT-5.6. In a memo, he told company employees that GPT-5.6 will be released first as a preview version, available only to a limited number of partners. The U.S. government will "approve customers' access one by one."
However, Ortermann also said, “We have clearly informed the U.S. government that this is not our preferred long-term approach. We will work with the government and other industry experts to develop more sustainable plans for future model releases.” In his memo, he added that he hopes to release the GPT-5.6 model “in a few weeks”.
According to sources cited by Axios News, Oltman met with U.S. Commerce Secretary Gina Rittner on the 24th local time. They discussed GPT-5.6, and Rittner expressed the hope that this model would undergo testing and approval by all relevant departments of the U.S. government.
Regarding this news, a White House official told CNN that the White House will continue to work with leading AI laboratories to develop joint solutions to address the challenges posed by AI technology. OpenAI declined to comment.
President Donald Trump signed an executive order on June 2, planning to establish a "voluntary review" framework that allows the U.S. government to obtain access to AI models up to 30 days before their release, enabling safety reviews in advance. This order is not mandatory, and companies can participate voluntarily.
On June 12 local time, the US government issued export control directives to Anthropic Company, ordering it to suspend access to Fable 5 and Mythos 5 models for all foreign citizens. Reuters reported that Rutnick's reason for this decision was “concern” that these models might be used by military intelligence agencies from so-called “US concern countries” such as China and Russia.
However, CNN noted that American AI regulation still lacks a transparent and unified framework. The export control directives regarding Anthropic come from the U.S. Department of Commerce, while those directed at OpenAI originate from the White House, which has confused American AI companies.
Brad Carson, the head of the bipartisan super political action committee ‘Public Priority’, believes: ‘The Fable incident shows that we need clear regulatory measures. Currently, we are adopting a temporary and opaque approach, which may even be illegal. It is certainly reasonable for the government to recall dangerous products including AI, but this must be done in a transparent and fair manner.’
