This week, President Trump signed an executive order that signals a meaningful shift in Washington’s approach to AI safety. The order asks AI companies to voluntarily submit their most powerful models for government testing up to 30 days before releasing them to the public. 

It also directs federal agencies to develop benchmarks to assess AI models’ cyber capabilities and create an AI cybersecurity clearinghouse to review and share information on vulnerabilities. The Alliance for Secure AI welcomes these steps. The admin is right that powerful AI systems demand federal attention before they reach the public.

However, we still need further steps to protect the American people.

The order itself states that “nothing in this section shall be construed to authorize the creation of a mandatory governmental licensing, pre-clearance, or permitting requirement for the development, publication, release, or distribution of new AI models”. 

This is exactly the problem. Ultimately, voluntary frameworks are only as strong as the goodwill of the companies participating in them. When the stakes are national security and economic stability, goodwill is not a sufficient safeguard.

Any developer unwilling to comply faces no legal consequence. That is a gap where bad actors and foreign competitors will exploit. Mandatory pre-release safety testing is not a burden on innovation; it is the baseline that any responsible industry standard demands.

America requires safety tests before a drug reaches pharmacies, before a plane carries passengers, and before a bridge opens to traffic. The same logic should apply to AI systems, which can disrupt critical infrastructure and democratic institutions. The President has done his part, and it is time for Congress to do theirs by making safety evaluations mandatory.