The Loyalty Doctrine: How Trump Reshapes Intelligence and AI Governance in His Image

On 2 June 2026, President Trump appointed Bill Pulte — a federal housing regulator with no intelligence or national security background — as acting director of national intelligence. Within hours of that announcement, Trump signed a revised executive order on artificial intelligence oversight, one that replaced mandatory pre-release government review of advanced models with a voluntary framework, after sustained industry objections. Two decisions, two days, one pattern.

The Pulte appointment is not anomalous. It is the latest expression of an administrative philosophy that subordinates competence to loyalty, and institutional memory to personal relationship. Pulte heads the Federal Housing Administration, a position he used to cultivate a public profile through online giveaways and social media engagement with Trumpworld. He is not a former intelligence official, a career diplomat, or a defence industry veteran. He is, by any conventional measure, an outsider to the machinery he has been asked to run. That he was chosen anyway tells us something structural about how this administration conceptualises intelligence itself — not as a repository of institutional knowledge, but as a tool of political management.

The role of director of national intelligence is supposed to coordinate the eighteen agencies that make up the US intelligence community, represent their assessments to the president and Congress, and act as an independent check on agency bias. It is a job that depends heavily on trust — trust from agency heads, trust from congressional oversight committees, trust from allied intelligence services who share classified material under the assumption it will be handled by professionals insulated from political interference. Placing a political loyalist in that role does not merely weaken the DNI's analytical capacity; it sends a signal to every intelligence professional that their assessments will be evaluated not on accuracy, but on alignment with the White House line.

This is not the first time Trump has installed loyalists in roles requiring independence. The pattern runs through his first term and into the second: acting secretaries who are either more ideological than their confirmed successors or more willing to implement politically motivated directives without the institutional resistance a Senate-confirmed official might mount. The DNI position, which by law can be filled on an acting basis indefinitely, is particularly suited to this strategy. No confirmation battle means no public scrutiny of Pulte's qualifications, no Senate hearing where career intelligence officers might surface concerns, and no paper trail of substantive questioning about his approach to, for example, Russia, Iran, or the intelligence community's role in domestic surveillance.

The AI executive order, signed the same day, follows a different but related logic. The original framework would have required frontier AI developers to submit their most advanced models to government review before public release — a form of pre-publication censorship, in the industry's framing, and a potential chokepoint for American AI competitiveness, in the tech lobby's argument. The revised order drops that requirement in favour of voluntary commitments. Companies developing models above a compute threshold are encouraged, not required, to notify the government and allow a review window. The enforcement mechanism is not penalties but reputational pressure: the hope that public disclosure of non-compliance will be sufficient deterrent.

This is, charitably, an optimistic theory of regulation. The companies that would most need oversight — those building sufficiently powerful models to pose national security risks — are precisely the companies with the most reputational capital to protect and the most resources to shape the narrative around voluntary compliance. Voluntary frameworks work when the incentive to comply is strong and the cost of non-compliance is credible. Neither condition is clearly met here. A company that declines to notify the government of a potentially dangerous model faces, under the revised order, the possibility of "public transparency" about that decision — which means a news cycle, not a civil penalty, not a criminal referral, not a contract suspension.

The tech industry's response to the original order reveals something about the political economy of AI governance. Companies that had spent years lobbying for light-touch regulation, that had argued against EU-style mandatory compliance frameworks on competitiveness grounds, found themselves confronting a White House that was — at least initially — more interventionist than expected. The response was swift and coordinated: trade group letters, op-eds from AI executives about American innovation being "chilled," and direct White House lobbying that produced the reversal within weeks. This is not a story about principled resistance to government overreach. It is a story about the relative ease with which a concentrated industry with direct access to the president can reshape regulatory frameworks in its favour.

The thread connecting the two decisions is the administration's broader approach to institutional authority. Intelligence agencies are not being staffed with intelligence professionals because the professionals represent an institutional perspective that may diverge from the president's political preferences. AI oversight is not being made mandatory because mandatory oversight implies the government has the right to evaluate private sector decisions, a principle that sits uneasily with an administration that sees the private sector — particularly its most politically connected corners — as a partner rather than a regulatee. In both cases, the logic is the same: institutions exist to serve the administration's political objectives, not to constrain them.

What remains uncertain is whether this approach produces functional failures or merely aesthetic ones. Intelligence communities are accustomed to political appointees; the institutional machinery is designed to absorb a certain amount of leadership churn and continue producing assessments. But the DNI's value lies precisely in being someone who can credibly tell the president what the agencies know, not what the president wants to hear. An acting director chosen for loyalty rather than credibility may still get briefings, but the nature of the conversation may shift — toward confirmation of existing assumptions rather than challenge of them.

The AI story is harder to evaluate in real time. No major safety failure from an unwatched frontier model has yet occurred; the risk is prospective. The voluntary framework may work if companies genuinely internalise safety norms and the reputational mechanism functions as intended. Or it may not, and the consequences — a model released publicly with latent capabilities that adversaries exploit — may become apparent only in retrospect, when the policy debate will be about a catastrophe rather than a near-miss.

Neither decision was made in isolation. They are part of an ongoing reorganisation of how the executive branch relates to the institutions it nominally directs — a reorganisation that favours flexibility over process, personal networks over career pipelines, and industry preferences over regulatory independence. The question is not whether this approach produces outcomes. It will. The question is which outcomes, for whom, and who bears the cost when the logic of loyalty collides with the logic of institutional function.

This publication covered the Pulte appointment through Reuters and the AI executive order through TechCrunch's reporting on the revised framework. Reuters led with the appointment's political dimension; TechCrunch focused on the industry lobbying that produced the reversal. The framing in this piece treats both stories as expressions of a single governing philosophy rather than isolated personnel decisions.