Commentary. The central question in Washington’s artificial-intelligence debate is not whether the United States should lead in AI. It is whether leadership requires rules strong enough to make the companies building the most powerful systems answerable for foreseeable harm.
Our view is that it does. Voluntary safeguards, company-led testing and outside audits may be useful, but they are not a sufficient public policy. When firms control the technology, define its risks and decide what evidence to disclose, the public is being asked to trust interested parties with a responsibility that belongs to democratic institutions.
The debate has become increasingly urgent. More than 1,500 AI-related bills were introduced in U.S. state legislatures in 2026, with 280 adopted, according to the National Conference of State Legislatures. The Trump administration has instead pursued a minimally burdensome national framework and has challenged state laws it considers inconsistent with federal policy. Claims Journal reports that the Justice Department created an AI Litigation Task Force in January to pursue those challenges.
There is a reasonable argument for federal consistency. A patchwork of state requirements could raise compliance costs, complicate interstate commerce and give the largest technology companies an advantage over smaller competitors that lack comparable legal teams. National standards could also prevent companies from moving development to jurisdictions with the weakest oversight.
But uniformity is not automatically accountability. A federal framework that merely asks companies to police themselves could pre-empt stronger state protections without replacing them with meaningful safeguards. That would simplify the industry’s legal environment while leaving workers, consumers and citizens with fewer remedies.
“The choice is not regulation or innovation. The choice is whether regulation is designed intelligently enough to protect both.”
The administration and its allies argue that aggressive rules could slow innovation and weaken the United States in its competition with China. That concern deserves to be taken seriously. AI development is global, and poorly designed licensing requirements or vague liability rules could push investment abroad. Existing criminal and civil laws may already cover some abuses, as Attorney General Todd Blanche has argued, particularly when people use AI to commit conventional crimes. IANS reported those arguments alongside calls from technology leaders and lawmakers for stronger safeguards.
Yet ordinary legal remedies are not enough for every frontier risk. They often arrive after damage has occurred, require identifiable victims and depend on evidence that may be difficult to obtain from proprietary systems. Preventive testing, incident reporting and independent evaluation are not attempts to ban innovation. They are ways to ensure that deployment does not become the experiment.
That does not mean every AI model needs the same regulatory burden. Rules should be proportionate to capability and use. A system that summarizes meeting notes should not face the same obligations as one used in critical infrastructure, medicine, financial decisions or autonomous cyber operations. Congress should establish clear thresholds, require documented safety testing for the most capable systems and preserve room for smaller developers to compete.
Independent evaluation is particularly important. Auditors should have access to enough information to test a model’s claims, while strong protections should guard legitimate trade secrets. Companies should also be required to report serious failures and retain records that allow regulators to investigate them. Without transparency, even well-written laws become difficult to enforce.
Critics will warn that regulators cannot keep pace with rapidly changing technology. They are right. But that is an argument for expert agencies, technical capacity and periodic review—not for abandoning public oversight. The same uncertainty that makes prediction difficult makes unchecked corporate discretion unacceptable.
Congress should therefore reject both extremes: a rigid system that freezes development and a voluntary regime that treats public safety as a branding exercise. The better path is a national baseline with enforceable duties, independent testing, targeted liability and room for states to address harms that federal rules overlook.
AI may deliver enormous benefits. That promise is precisely why its governance cannot be reduced to a race for speed. Innovation earns public legitimacy when the people who profit from it also accept clear responsibility for its consequences.