USA introduces new state review regime for the most powerful AI models

The White House will hold a meeting with technology sector leaders to discuss a new voluntary state review system for the most powerful artificial intelligence (AI) models. According to an American administration official, the initiative aims to establish a mechanism for assessing the risks associated with new technologies.

Regulations and Verification Mechanism

The new regulatory regime is defined in an executive order signed by President Donald Trump on June 2. According to the provisions, companies such as OpenAI, Google, and Anthropic will be able to provide the government with access to their most powerful models for a period of 30 days prior to their planned public release.

This step was taken following concerns regarding Anthropic's Mythos model. The company initially refused public access to it due to its ability to identify vulnerabilities in the computer systems of banks, governments, and hospitals. Ultimately, the model was released to the market, but with imposed restrictions on its use.

National Security and Previous Incidents

Previous cybersecurity issues have already forced authorities to intervene. In June, the US Department of Commerce ordered Anthropic to cease all foreign access to the Claude Fable 5 and Mythos 5 models. The reason was national security concerns, after a case of users bypassing built-in safeguards was reported.

Following the incident, the company was forced to withdraw both models for more than two weeks before access to them was gradually restored.

Shift in Administrative Policy

The introduction of the assessment regime marks a significant turnaround in the Trump administration's approach. Until now, the government applied a policy of non-interference, arguing that the state should not hinder innovation within the global race with China for technological leadership in the field of artificial intelligence.