An America First Federal AI Framework: Recommendations to Build on Executive Order 14365

Overview

President Donald J. Trump has demonstrated historic leadership in positioning America to win the global race for artificial intelligence (AI) dominance. From revoking the Biden Administration's burdensome AI executive order to unveiling an ambitious AI Action Plan and the Stargate infrastructure initiative, this Administration has made clear that American AI leadership is a national priority.

The America First Policy Institute (AFPI) holds that Executive Order (EO) 14365, Ensuring a National Policy Framework for Artificial Intelligence, is an important effort that builds on the Trump Administration’s leadership on this issue to date. The EO introduces a unified national approach to AI governance in a way that prioritizes American families and workers while promoting the innovation that keeps America ahead of China.

Since its founding, the United States has struck a dynamic balance between state and federal governance. States have successfully played major roles in governing numerous industries, including automobiles and insurance. Yet, the Founders also understood that sometimes unified federal law is required, with interstate commerce and national security as two prominent examples. The executive branch should therefore be discerning when imposing constraints on state lawmaking, in turn carefully targeting litigation, funding conditions, and other measures at only the most egregious state excesses.

The EO identifies the threat posed by state laws that mandate ideological outcomes in AI systems because such laws impose significant compliance burdens on developers while incentivizing biased model outputs to satisfy disparate impact standards. Executive Order 14365 directs the Secretary of Commerce to evaluate state AI laws and identify those that conflict with American AI leadership. Federal action against state laws should be judicious and exacting, targeting the worst excesses rather than sweeping away legitimate state protections. States like Utah and Arkansas have enacted transparency and child safety measures that protect families without mandating ideological outcomes. Federal standards should build on these state initiatives, establishing a framework that protects children, prevents censorship, respects copyrights, and ensures all Americans benefit from such protections.

The executive order also correctly notes that certain state AI laws, like those of Colorado, pose a direct threat to American AI leadership because they compel embedding anti-American ideologies like diversity, equity, and inclusion (DEI) into AI systems. These woke mandates force companies to produce outputs based on predetermined ideological outcomes rather than truth. Such laws, based on the radical progressive agenda of the Left, are antithetical to American values. Furthermore, they impose unreasonable burdens on American innovators, fragmenting the regulatory landscape and ceding ground to foreign competitors like China.

State AI laws vary in their effects on innovation and free expression. Laws that mandate ideological outcomes differ in kind from laws that give the public more information or protect families. Transparency requirements do not tell AI companies what values to embody or what outputs to produce. They let Americans see for themselves what choices companies have made and judge accordingly. Informed consumers can then reward companies that share their values while avoiding those that do not.

The executive order targets laws that “require AI models to alter their truthful outputs.” This is the appropriate target for combatting onerous regulation. Transparency requirements that inform the public without compelling speech or mandating ideological conformity are consistent with American values. They should be preserved, or better, strengthened through a federal standard.

Recommendations for a Uniform Federal Framework for AI

The President has wisely called on Congress to develop legislation establishing a uniform federal policy framework. Executive Order 14365 identifies key priorities: protecting children, preventing censorship, respecting copyrights, and safeguarding communities. Additionally, a federal AI framework must build AI readiness and expertise in government to better understand AI’s developments and opportunities. AFPI offers the following four categories of recommendations drawn from our America First AI Agenda:

(1) Child Safety Standards

Few things are more important than the sanctity of the American family. The federal framework should consider the following recommendations:

•  Implement age verification requirements. Provide guidance on technical methods for chatbot and app store providers to develop and enforce age verification systems based on technical best practices.
• Require reporting of dangerous interactions. Require companies to detect dangerous or prohibited conversations between minors and chatbots and report them to parents and appropriate authorities.
• Prohibit harmful content to minors. Prohibit AI chatbots from engaging in conversations related to self-harm, suicide, and sexual content with minors through refusal training, age verification, and other methods.
• Ban AI-generated Child Sexual Abuse Material (CSAM). Prohibit the creation, distribution, and possession of all AI-generated sexually explicit images of minors.
  

(2) Transparency and Public Accountability

Americans deserve to know what values are being embedded in AI systems they encounter at work, in their children’s classrooms, and through public services. To address these issues, the federal framework should consider these recommendations:

• Require transparency on AI values and security. Require AI companies to publish information about subjective judgments embedded into model behavior, as well as significant security incidents and developer evaluations of model capabilities with potential risks to national security.
• Evaluate AI for false and deceptive outputs. Conduct evaluations of false, unreliable, or deceptive behavior of domestic and foreign AI, including woke ideology and CCP propaganda, and publish results to inform federal procurement decisions and public awareness.
• Develop guidance on reliable, non-deceptive AI. Provide concrete evaluation baselines for developers to identify and correct deceptive behavior.
• Mandate disclosure of political AI compensation. Require AI companies to publicly disclose compensation received for services relating to how the system presents information about candidates, campaigns, or ballot measures.

(3) Standing Up for the American Worker

AI will change how work is done, but that does not mean leaving hardworking Americans behind. To protect the American worker, the federal framework should consider these recommendations:

• Track AI’s impact on skill demand. Develop and incorporate AI-specific metrics in surveys of job and labor market data.
• Train Americans in high-demand skills. Expand programs in AI-impacted sectors with a focus on vocational training and registered apprenticeships.
• Promote reliability in AI agents. Develop voluntary guidance on reliability, attributability, and trustworthiness of AI agents to ensure they serve as partners to American workers rather than unreliable replacements.

(4) Building AI Readiness in Government

The federal government needs a nerve center of AI excellence that can monitor emerging capabilities, assess their implications for national security, and coordinate responses across agencies. To reinforce AI readiness in government, the federal framework should consider these recommendations:

• Authorize and fund the Center for AI Standards and Innovation (CAISI). CAISI is critical for ensuring that the federal government has strategic awareness of AI.
• Focus CAISI’s mission on innovation and national security. Ensure that CAISI’s mission adheres to the following: promoting and protecting U.S. AI leadership, measuring AI capabilities relevant for national security, and issuing voluntary guidance on AI security.
• Expand hiring authorities for AI-related positions across government. Tools such as Direct Hire Authority and Excepted Service Hiring could be leveraged for top talent to enter and exit public service faster and with less red tape.

Conclusion

AFPI applauds the issuance of Executive Order 14365 and shares the Trump Administration’s vision of American AI dominance achieved through a unified national framework. AFPI supports immediate action against woke state AI mandates that threaten innovation and encourages Congress to pursue comprehensive federal framework that protects American families, prevents censorship, and unleashes American AI leadership. That is the America First approach.