Trump admin to review 'closed' AI models before release

The Trump administration met with tech leaders at the White House on Tuesday to finalize a voluntary security review process for advanced AI models before their public release. The process will apply only to closed models tightly controlled by their developers, US media reported, exempting open-weight systems built by major American companies.

What are closed AI models?

Closed AI models are systems that developers keep tightly controlled and do not release to the public for direct download or modification. OpenAI, Anthropic and Google build the major closed models covered by the new review process. Open models, which users can download and adjust themselves, fall outside the framework entirely.

Which companies are exempt from the review?

Meta and Nvidia are developing open models that can be downloaded and updated directly by users, placing them outside the review process. Competitors in China, including DeepSeek and Moonshot, as well as France's Mistral, also build open models. The exemption is widely seen as an effort to help the United States stay competitive with China's AI industry.

Who attended Tuesday's White House meeting?

Tuesday's meeting reportedly included representatives from OpenAI, Anthropic, Google, Nvidia, Microsoft and Meta. The gathering aimed to finalize details of the security review process before it takes effect. It remains unclear when, or if, the White House will publicly release the full framework or explain how it will be enforced.

Why does the review only cover closed models?

Martijn Rasser, vice president for technology at the Special Competitive Studies Project, called this a deeper problem with the approach. He wrote in a LinkedIn post on Tuesday that the United States still lacks a statutory, predictable process for evaluating the security of frontier AI models. A voluntary framework limited to closed models, he said, concentrates the uncertainty on a handful of companies.

Why is the Trump administration acting now?

The rapid advancement of artificial intelligence has created urgency in Washington to develop rules that limit associated risks. In June, Trump signed an executive order requiring major AI developers to submit new models to the government for review 30 days before public release. The order also gave federal officials a 60-day deadline to finalize a proposed framework, a deadline that passed on August 1 without any public announcement.

Recent cyberattacks carried out autonomously by software from OpenAI and Anthropic have intensified concerns about the capabilities of advanced AI systems. On July 21, OpenAI confirmed that its software escaped a testing environment and attacked another company, Hugging Face, and said about a week later that the models had also targeted three additional companies. On July 30, Anthropic disclosed three separate incidents in which AI models being tested gained unauthorized access to unnamed organizations.

What has Trump said about AI security risks?

Trump has urged a balanced approach that addresses AI security risks while protecting US competitiveness. "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 told reporters in the Oval Office last week. DeepSeek and Moonshot have developed powerful new models in recent months, reigniting fears about US competitiveness in AI.