President Donald Trump’s administration scored a quiet diplomatic win this week. G20 innovation ministers reached consensus on the Carolina Principles for Emerging Technologies, endorsing pro-innovation policy frameworks, foundational research, commercialization, and trusted deployment. The instinct is sound: Washington should not treat every new technology as an excuse to build another regulator.
But as government itself adopts AI, there is a prior democratic question: If voters can replace an administration, can its successor also replace the automated policies it leaves behind?
Trump’s comeback is a case study in American democracy’s greatest strength: political direction can be reversed. In 2016, Americans sent an outsider to Washington to disrupt a political and administrative establishment they no longer trusted. In 2020, they removed him. In 2024, they returned him to office with a new mandate. The sequence was bruising and loud, but it was constitutional self-government in motion: a president could be replaced, his successor could govern, and voters could reverse the reversal.
Trump’s first term made unelected administrative power a national political question. His 2017 regulatory order directed most agencies to identify at least two existing rules for repeal for every new one. He appointed three Supreme Court justices, and all three later joined the Loper Bright majority that ended mandatory Chevron deference and restored courts’ independent judgment over agencies’ readings of federal law. These were different actions, but they expressed the same constitutional instinct: delegated power must remain bounded by accountable institutions.
Trump’s second term has pressed that principle more directly. Executive Order 14171 reinstated a classification intended to make policy-influencing civil-service positions more answerable to elected executive leadership. Executive Order 14192 expanded the regulatory offset from two-for-one to 10-for-one. And America’s AI Action Plan made rapid adoption, infrastructure, and global AI leadership central administration priorities. One can dispute individual implementations while recognizing the democratic premise: Public power should not become immune to electoral correction.
The administration’s own OMB Memorandum M-25-21 does more to honor that premise than critics acknowledge. It accelerates federal AI use while requiring predeployment testing, impact assessments, signed risk acceptance, monitoring, human oversight, appeals, and the discontinuation of noncompliant high-impact uses. That is risk-based governance without a new super-regulator.
But the framework has a blind spot: democratic succession. A human official leaves office. A model embedded in an agency workflow can retain its thresholds, categories, data assumptions, vendor constraints, and policy choices. An incoming secretary may reverse a memorandum while unknowingly inheriting hundreds of machine-mediated decisions. Formal policy changes; operational policy survives in code.
That would weaken the very feature that Trump’s two terms have showcased: reversibility. America’s democratic advantage over authoritarian systems is not flawless decision-making. It is that citizens can challenge a decision, remove the decision-maker, and authorize a different course.
The fix should be modest. OMB should add a one-page democratic-transition annex to the impact assessment already required for every high-impact federal AI use. It should identify the statute authorizing the function, the decisions the system influences, the material policy judgments embedded in its design, the accountable official, any vendor dependencies, and the procedure for suspending or replacing the system without the vendor’s permission.
Within 120 days of taking office, each incoming agency head should affirm, amend, or retire those machine delegations. Systems necessary for continuity may keep operating during review, but their authority should not expand. OMB can establish the process now by extending the documentation agencies already produce. Congress should later codify its essentials so no administration can erase democratic reversibility with a memorandum.
This is not anti-AI. It matches Trump’s preference for innovation without unnecessary bureaucracy and his insistence that policy-driving power answer to elected leadership. It also reflects the oldest American rule of public authority: Power is delegated, never surrendered.
Trump was elected, defeated, and elected again. Courts reclaimed independent judgment over agencies. Presidents reversed their predecessors. That is not democratic failure. It is proof that the American system can absorb conflict without surrendering the people’s right to change direction.
TRUMP PROMISED TO DRAIN THE SWAMP. AI IS MAKING IT UNFIREABLE
The government’s machine layer should be built in the same spirit. A policy that cannot be discovered, contested, or reversed by the next lawful administration is no longer merely a tool; it is a small pocket of government insulated from elections.
Trump has spent two terms making Washington more answerable to voters. His most durable AI legacy could be ensuring that America leads in artificial intelligence without surrendering its democratic capacity to change its mind.
Burak Oktenli is an independent researcher based in Washington, D.C. He holds a bachelor’s degree in computer science and engineering from the University of South Florida and an MBA, and is completing a master’s of professional studies in applied intelligence at Georgetown University. His writing focuses on defense strategy, alliance burden-sharing, military technology, and the governance of autonomous systems.