Anthropic’s Claude Mythos—a frontier AI model believed to have uncovered thousands of security flaws in critical digital infrastructure—has prompted the White House to take a harder line on AI guardrails.
Several reports indicate that the administration is considering safety testing for federal contractors to address the threats AI systems pose to national security and the financial sector.
New polling from the Institute for Family Studies (IFS) should encourage the White House to take even stronger measures. The survey found that 82% of Americans support mandatory AI safety testing, and Trump voters polled even higher, with 90% of them in favor.
The same poll found that 88% of Americans believe AI systems should be tested for national security risks, and 87% favor safety testing against broader risks for children and families.
This is a mandate. To deliver on it, President Donald Trump should leverage the federal government’s purchasing power and consumer protection capabilities to secure AI safety standards that put American children, parents, workers, and communities first.
American families and communities need a voice in AI governance.
Large data center projects are jacking up electricity bills and generating an enormous grassroots backlash. AI “companions” and chatbots are inducing kids to suicide, helping to plan mass shootings, and fracturing American family life.
Big Tech CEOs like Dario Amodei and Sam Altman prophesy of a dystopian future where AI replaces all human jobs, “knows” you better than you know yourself, and runs everything from your household to the government.
Unsurprisingly, Americans are concerned about the messaging from Big Tech, and they strongly favor AI safety standards with teeth.
Beyond national security and finance, the IFS polling identified eight additional risks that overwhelming majorities of both Trump and Harris voters want. Key areas of concern include AI’s risks to mental health, kids’ safety, job security, human privacy, and basic decision-making.
While the Center for AI Standards and Innovation (CAISI) within the U.S. Commerce Department already vets some AI models for security-related issues, its current mandate does not prioritize family and community concerns.
A Presidential Commission on AI and the American family would remedy that by empowering community and family voices to surface legitimate issues and channel them into actionable guidance for AI safety standards.
President Trump should direct CAISI to develop standards and protocols for AI safety testing based on guidance from the presidential AI commission.
Safety measures should evaluate whether frontier AI models display sycophantic or manipulative tendencies that disproportionately harm vulnerable populations like kids, the mentally disabled, and the elderly. Such standards should also test the implementation and effectiveness of AI safeguards in mitigating harmful outcomes.
The president should also direct other agencies, like the Federal Trade Commission (FTC) and the Departments of Labor, Energy, and Education to develop a broader suite of best practices, audits, and reporting standards for high-impact AI developers.
These measures would draw on the AI commission’s guidance to help identify and address other AI risks in job security, human decision-making, and privacy. For example, testing standards could focus on the effects of AI educational technology on student privacy and learning outcomes.
To ensure AI evaluation and safety standards keep pace with rapid technological change, relevant federal departments, in consultation with the AI commission, could designate and resource private independent verification organizations (IVOs).
IVOs would help translate positive and negative outcomes identified by community stakeholders into safety standards. These new entities would perform ongoing evaluations before and after deployment to ensure that tech companies remain accountable to the American people.
But creating standards and verifying implementation by frontier AI firms is not enough.
AI companies only provide transparency on a limited and voluntary basis. Real accountability will require leveraging federal AI procurement policy, consumer protection authority, and other tools to ensure AI systems are made to put families and communities first.
On the procurement side, President Trump should direct the Office of Management and Budget to require AI companies to participate in ongoing safety testing and reporting as a condition of receiving government contracts.
Although federal agencies are barred by executive order from procuring ideologically biased AI systems, the president should go further by requiring Big Tech companies to cooperate with IVO safety testing and evaluations for public-facing AI systems.
At the same time, President Trump should direct the FTC to review its regulatory rulemaking and guidance on unfair and deceptive trade practices to address risks surfaced by the AI commission. Revised guidance from the FTC could provide strong incentives for AI developers and deployers to comply with AI safety protocols.
Finally, the president should order a comprehensive review of all federal government benefits and subsidies provided to Big Tech companies. AI companies that benefit from expedited permitting, federal land use, tax abatements, or loan guarantees should be required to operate in the public interest.
President Trump has a golden opportunity to fulfill the wishes of his supporters and deliver trustworthy AI for American families and communities.
He must act decisively to ensure that AI development prioritizes the common good over Silicon Valley elites.
We publish a variety of perspectives. Nothing written here is to be construed as representing the views of the Daily Signal.
