The Trump administration has quietly changed course on artificial intelligence oversight. A new executive order replaces mandatory reporting requirements with a voluntary framework for companies to share their AI models with the government. The shift marks a significant departure from earlier regulatory efforts and has drawn attention from both industry insiders and security experts.
Voluntary model sharing replaces earlier mandates
Under the new order, developers of advanced AI systems are no longer required to submit their models for government review. Instead, they can choose to share details of their systems on a voluntary basis. The change effectively removes the threat of penalties for non-compliance that existed under previous directives.
The administration says the voluntary approach is meant to encourage cooperation rather than force it. Companies that opt in will provide security reviewers with access to model architectures, training data, and deployment safeguards. The goal is to identify potential risks before systems reach the public.
Security review implications
The voluntary framework could still enhance national security reviews, according to officials familiar with the order. If a critical vulnerability is discovered during a voluntary review, the government would be able to work with the company to patch it before any harm occurs. But critics question whether companies will actually participate without a legal obligation.
Security experts have long warned that rapid AI deployment outpaces traditional oversight. The voluntary model leaves a gap: there is no mechanism to compel sharing from firms that choose to stay silent. Some inside the administration argue that transparency efforts will still benefit from goodwill, while others worry that bad actors will simply opt out.
Unresolved questions on regulatory approach and competitiveness
The shift raises unresolved questions about the future of AI regulation in the United States. Without mandatory reporting, the government may lack a comprehensive picture of what advanced models are being built. That could leave critical blind spots if a dangerous capability emerges in a proprietary system.
There is also the matter of global competitiveness. The United States is racing against other nations, notably China, to lead in AI development. Some policymakers argue that heavy-handed regulation would stifle innovation and drive companies overseas. Others contend that the new voluntary framework is too weak to address real risks and that the U.S. risks falling behind on safety standards.
The executive order does not set a deadline for companies to decide whether to participate. It also does not specify what, if any, public reporting will occur on the models that are voluntarily shared. For now, the government is waiting to see how many firms step forward.
