Trump Shelves AI Security Review Executive Order

U.S. President Trump decided on Thursday to temporarily halt the signing of an artificial intelligence executive order that was originally set to take effect the same day. The order aimed to establish a government security review mechanism before the release of AI models. This move is a significant relief for tech companies, but the policy tussle surrounding AI regulation might have just begun.

It is reported that the aforementioned executive order planned to set up a voluntary review process, requiring AI companies to submit new models for review to federal agencies prior to release. This would allow the government to assess potential national security risks and buy time to protect its own network security. Recently, companies such as Anthropic have granted early access to their latest models to some partner institutions to help fix potential security vulnerabilities that the models might expose. The executive order was an attempt to formalize this practice.

However, Trump expressed to reporters in the Oval Office of the White House that he was concerned that a formal review procedure would slow down the development pace of American AI companies and put them at a disadvantage in the technological competition with China. "We are leading China, leading everyone, and I don't want to do anything that harms this lead," Trump said, "I really think it could become a stumbling block."

This marks another reversal in AI administrative regulatory policy. Former President Biden had signed an executive order during his tenure, requiring AI companies to share some safety details and test results of models with the government, but that order was revoked in the early stages of Trump's inauguration. When asked whether he would sign a new order after revision, the White House directed reporters to Trump's aforementioned public statement without further explanation.

In the meantime, federal, state, and local governments in the United States are facing increasingly complex pressures in AI governance. Some local governments have slowed down the approval of AI data centers due to public opposition; several states have passed or are considering special legislation involving AI applications, including regulations on the use of AI in recruitment; and various polls also show that there is widespread distrust among the American public towards AI technology.

The shelving of this executive order makes the direction of the U.S. AI regulatory framework even more unclear. Analysts point out that in the absence of federal rules, the fragmented legislation at the state and local levels may further accelerate, and the compliance uncertainty faced by businesses will be difficult to eliminate in the short term.