Trump Says America Must Win AI. Who Decides Whether It Is Safe?
President Donald Trump’s dismissal of AI safety warnings has sharpened a question Washington cannot settle through confidence alone: who should determine whether increasingly powerful systems are ready for widespread use, and what evidence should that decision require?
Reuters reported on September 14 that Trump rejected demands for additional oversight and argued that existing criminal and regulatory powers were sufficient. His position places the burden on critics to demonstrate why current arrangements cannot manage emerging risks.
The China argument carries weight—and has limits
Trump’s strategic concern is clear. Speaking in Ireland on September 13, he emphasised preserving America’s lead over China and questioned warnings about AI’s development.
That position fits the administration’s broader programme. Its July 2025 AI Action Plan organised policy around innovation, infrastructure and international leadership, treating technological expansion as a source of national power.
The argument deserves serious consideration: poorly designed rules could delay useful development or impose costs that smaller competitors struggle to bear. Yet competition cannot answer whether a particular system has been adequately tested. A geopolitical objective and a safety judgment require different evidence.
Existing authority is not the same as demonstrated readiness
The United States already has an institutional basis for assessing AI risk. NIST’s voluntary AI Risk Management Framework addresses trustworthiness across design, development, use and evaluation. NIST says the framework is being revised and has also begun work on a critical-infrastructure profile.
That complicates any claim that government must choose between unrestricted development and abandoning innovation. The practical debate concerns which safeguards should remain voluntary, where independent scrutiny is necessary and who has the expertise to evaluate a company’s assurances.
Public confidence requires more than pointing to enforcement powers. Policymakers should explain how they would identify a dangerous capability, assess its significance and act before deployment creates consequences that are difficult to reverse.
Industry expertise needs independent scrutiny
Developers possess knowledge governments need, but they also have commercial interests. Their safety warnings deserve examination; their preferred regulatory designs deserve examination too.
Demanding expensive compliance without considering proportionality could strengthen established firms. Leaving evaluation entirely to developers could allow the organisations benefiting from deployment to define acceptable risk. Neither outcome should be accepted without scrutiny.
A more credible debate would specify the systems being assessed, the evidence required and the conditions triggering intervention. Catastrophic scenarios should be distinguished from demonstrated harms and measurable vulnerabilities. Uncertainty warrants investigation; it does not justify presenting either disaster or safety as guaranteed.
WARYATV Assessment
America’s AI advantage will be more durable if institutions can test the claims supporting it. Trump’s approach becomes vulnerable if confidence in leadership substitutes for demonstrable safeguards. Congress and regulators should demand evidence that the public can evaluate. The decisive measure of leadership is whether technological progress produces dependable capability while preserving accountability for its consequences.
