US requests exclusive government access to AI models 30 days before release

The Trump administration is asking AI companies for early and exclusive access to new, advanced AI models. The US government wants this access 30 days before other companies and organizations. This serves to assess and mitigate security risks.

US President Donald Trump signed a decree yesterday with this request to AI companies. The presidential decision was only expected to come last month and in the initial version asked for a priority period of 90 days, CNN writes. The official reason for this delay was that President Trump did not like certain aspects of the decree. According to him, it would be an obstacle to AI development. Those concerns seem to have disappeared now that the signing has been done.

Formally, it is a request for AI companies to voluntarily grant access to their new AI models to the United States government. The period of 30 days of exclusive access should allow the government to identify risks to limit their impact. It is not clear what the consequences will be if AI companies refuse. This concerns AI models that have advanced cyber capabilities, such as scanning for vulnerabilities in IT systems.

In private

Examples of such AI models are Mythos from Anthropic and GPT-5.5-Cyber from OpenAI. AI company Anthropic is in a conflict with the Trump administration over ethical aspects of AI use. Yet the US government is among the select group of agencies and companies that were given exclusive access to Mythos by Anthropic. Yesterday, the company announced that it is expanding this private access to 150 organizations in 15 countries.