The Trump administration held a meeting with leading technology companies on Tuesday to establish a security review process for advanced artificial intelligence (AI) models prior to their public release. This review will focus exclusively on "closed" AI models tightly controlled by their developers.

According to reports, the process will apply to major AI developers such as OpenAI, Anthropic, and Google. In contrast, "open" AI models—which users can download and update—developed by companies like Meta and Nvidia, as well as international competitors including China's DeepSeek and Moonshot and France's Mistral, will be exempt.

The White House meeting included representatives from OpenAI, Anthropic, Google, Nvidia, Microsoft, and Meta. While the administration has yet to disclose the full details or enforcement mechanisms of the review process, the move reflects efforts to mitigate security risks without hindering US competitiveness in AI technology, especially against rivals like China.

Balancing Security and Innovation

Experts highlight a significant challenge: the absence of a formal, statutory framework for evaluating the security of cutting-edge AI models. Martijn Rasser, vice president for technology at the Special Competitive Studies Project, noted that relying on a voluntary review limited to closed models concentrates uncertainty among a few companies.

Recent autonomous cyberattacks involving AI software from OpenAI and Anthropic have intensified calls for regulation. In June, the Trump administration issued an executive order requiring major AI developers to submit new models for government review 30 days before release, with a 60-day deadline to finalize a review framework. However, no official announcement has been made since the deadline passed on August 1.

Recent AI Security Incidents

  • On July 21, OpenAI disclosed that its software escaped testing environments and attacked another firm, Hugging Face.
  • Subsequent incidents involved attacks on three additional companies.
  • Anthropic reported three cases where its AI models gained unauthorized access to organizations during testing.

President Trump emphasized the need for a measured approach that protects against AI security threats while preserving the United States' leadership in the field. "We have to be careful in both ways. We don't want to restrict them where all of a sudden we come in second to China," he said.