AI Revolving Door Analysis

Quick Assessment

Key Links

Overview

The AI revolving door describes the movement of personnel between frontier AI laboratories, government regulatory bodies, safety institutes, think tanks, and academic institutions. The phenomenon encompasses multiple directional flows: AI researchers and executives entering government advisory or regulatory roles; former government officials joining AI companies; think tank personnel cycling through White House advisory positions; and safety researchers migrating between independent organizations and commercial labs. As AI governance has become a high-stakes policy domain, the velocity and visibility of these transitions have accelerated sharply, particularly since 2022.

The pattern mirrors the well-documented revolving door in financial services and healthcare but carries distinct features. Unlike banking, where regulatory expertise is relatively transferable across institutions, AI policy expertise is often lab-specific — a person who understands OpenAIOrganizationOpenAIComprehensive organizational profile of OpenAI documenting evolution from 2015 non-profit to Public Benefit Corporation, with detailed analysis of governance crisis, 2024-2025 ownership restructuri...Quality: 62/100's internal safety processes carries knowledge that is directly valuable to regulators and competitors alike. This creates both a rationale for personnel flows (genuine expertise transfer) and a heightened capture risk (access to information that can be leveraged for competitive or regulatory advantage).

Research on revolving doors in adjacent sectors provides the most rigorous quantitative evidence available. Studies of USPTO patent examiners find that those planning to enter private employment grant 12.6–17.6% more patents to prospective employers, with the patents receiving fewer subsequent citations — indicating quality degradation alongside favoritism.¹ A study of 420,153 top corporate executives across 12,869 firms found that more than half had prior executive branch experience, and firms were significantly more likely to win government procurement contracts within two years of hiring a former agency official.² These findings have not been replicated specifically for AI governance, but they frame the prior for assessing AI-sector dynamics.

Conceptual Framework

Directional Flow Categories

The AI revolving door can be decomposed into five recurring movement patterns:

Lab-to-government flows occur when AI company personnel — researchers, policy staff, or executives — move into regulatory agencies, safety institutes, or White House advisory roles. These transitions are often framed as public service contributions, bringing technical knowledge into institutions that might otherwise lack it. The risk is that the individual retains informal loyalty to, or future employment interest in, their former employer.

Government-to-lab flows describe former officials, military leaders, or elected members joining AI companies' boards, policy teams, or advisory structures. These moves typically bring government access and credibility, raising concerns about whether the official's prior regulatory decisions were influenced by anticipation of private-sector opportunities.

Think tank cycles involve researchers at policy institutes (CSET, CNAS, RAND, etc.) rotating into White House, NSC, or Congressional staff roles, then returning to think tanks or moving to industry. Think tanks occupy a structurally ambiguous position: they are nominally independent but often funded by technology companies and foundations with AI investments.

Academia-to-industry flows describe university researchers — particularly those who built foundational AI capabilities — transitioning into commercial roles, sometimes while retaining academic appointments. This flow raises questions about whether academic prestige launders commercial interests.

Safety organization cross-staffing involves personnel moving among AI safety nonprofits (MIRIOrganizationMachine Intelligence Research InstituteComprehensive organizational history documenting MIRI's trajectory from pioneering AI safety research (2000-2020) to policy advocacy after acknowledging research failure, with detailed financial da...Quality: 50/100, FHIOrganizationFuture of Humanity InstituteThe Future of Humanity Institute (2005-2024) was a pioneering Oxford research center that founded existential risk studies and AI alignment research, growing from 3 to ~50 researchers and receiving...Quality: 51/100, ARCOrganizationAlignment Research CenterComprehensive reference page on ARC (Alignment Research Center), covering its evolution from a dual theory/evals organization to ARC Theory (3 permanent researchers) plus the METR spin-out (Decembe...Quality: 57/100), safety teams at frontier labs (OpenAIOrganizationOpenAIComprehensive organizational profile of OpenAI documenting evolution from 2015 non-profit to Public Benefit Corporation, with detailed analysis of governance crisis, 2024-2025 ownership restructuri...Quality: 62/100, AnthropicOrganizationAnthropicComprehensive reference page on Anthropic covering financials ($380B valuation, $14B ARR at Series G growing to $19B by March 2026), safety research (Constitutional AI, mechanistic interpretability...Quality: 74/100, Google DeepMindOrganizationGoogle DeepMindComprehensive overview of DeepMind's history, achievements (AlphaGo, AlphaFold with 200M+ protein structures), and 2023 merger with Google Brain. Documents racing dynamics with OpenAI and new Front...Quality: 37/100), and government AI safety institutes (UK AISI, US AISI). These flows are particularly consequential for AI safety because they determine where alignment expertise is concentrated and whether safety-oriented researchers can maintain independence from commercial pressures.

Mechanisms of Influence

Personnel flows exert policy influence through several distinct mechanisms:

• Prospective employer favoritism: Officials anticipating private-sector employment may grant favorable treatment to prospective employers — the mechanism documented in patent examiner research.¹
• Access deployment: Former officials join companies specifically to leverage ongoing relationships with former colleagues, accelerating regulatory access beyond what formal lobbying provides.
• Information asymmetry: Officials who move from government to industry carry institutional knowledge — internal deliberative processes, unpublished policy directions — that confers competitive advantage.
• Agenda-setting: Former industry personnel in government shape which questions get asked and which evidence gets weighted, without necessarily acting in bad faith.
• Research influence: Industry funding of think tanks and academic institutions creates soft dependence that shapes research priorities before personnel transitions even occur.³

Named Moves: Documented Transitions

The following table compiles documented personnel transitions relevant to AI governance. Dates derive from the research data; where the research provides only approximate periods, this is noted.

Note: MIRI-to-lab and FHI-to-lab flows are a documented pattern in the safety research community. FHIOrganizationFuture of Humanity InstituteThe Future of Humanity Institute (2005-2024) was a pioneering Oxford research center that founded existential risk studies and AI alignment research, growing from 3 to ~50 researchers and receiving...Quality: 51/100 closure in 2024 accelerated researcher dispersal; a significant fraction moved to frontier lab safety teams or to AnthropicOrganizationAnthropicComprehensive reference page on Anthropic covering financials ($380B valuation, $14B ARR at Series G growing to $19B by March 2026), safety research (Constitutional AI, mechanistic interpretability...Quality: 74/100. Specific named transitions are not fully documented in the available research data and are omitted to avoid hallucination.

AISI Cross-Staffing

The AI Safety Institutes established in the UK and US following the 2023 Bletchley Park summit represent a novel institutional structure with built-in revolving door dynamics. The UK AISI was designed to recruit technical staff with lab experience — its evaluation mandate requires people who understand how frontier models are built. The Jade Leung transition exemplifies this: genuine expertise transfer from a lab's policy function to a government evaluation body. The governance question is whether such staff retain commitments or access relationships that compromise independent assessment.

The US AISI (housed within NIST) and analogous bodies in Canada and Japan have pursued similar recruitment strategies, drawing on researchers from academic AI safety communities and, to a lesser extent, from within labs. Cross-staffing between national AISIs — through joint evaluations and staff exchanges — creates a further layer of personnel flows with less public visibility than the lab-to-government transitions covered by standard lobbying disclosure requirements.

Quantitative Analysis

Comparative Evidence from Adjacent Sectors

Direct quantitative studies of the AI revolving door are not yet available. The following table summarizes the best available evidence from structurally similar sectors:

AI Lobbying Trajectory

The scale of AI-sector political influence has grown sharply. Lobbying reports mentioning AI doubled from 2022 to 2023, then nearly doubled again in 2024.⁷ By 2025, more than 3,500 lobbyists — representing approximately 25% of all federal lobbyists — reported lobbying on AI issues.⁷ The broader technology sector ranks as the third-largest employer of former government workers turned lobbyists, behind only healthcare and financial services.⁸

These figures frame the revolving door not as an incidental feature of AI governance but as an organized influence strategy. The Revolving Door ProjectOrganizationRevolving Door ProjectGovernment accountability project (of the Goodnation Foundation) led by Executive Director Jeff Hauser. Funded by Democracy Fund. Maintains the "Tracking Uses of AI in the Trump Administration" tra... has documented industry funding of policy groups — including organizations promoting AI and data center expansion — as part of this broader ecosystem.⁹

Government AI Deployment Growth

A Washington Post analysis found approximately 3,000 AI deployments across 29 federal agencies, up 75% from the prior year.⁹ This expansion occurred largely under the oversight of officials with prior industry ties, raising questions about whether procurement decisions reflected public interest or cultivated relationships.

Strategic Importance

For AI Governance

The revolving door is among the most tractable structural risks in AI governance. Unlike technical alignment problems, it is a known institutional phenomenon with documented regulatory interventions — cooling-off periods, disclosure requirements, recusal rules — even if current implementations are inadequate. The AI safety community and policymakers who want stronger oversight face a structural disadvantage: government AI safety capacity is built largely by recruiting from the industry it regulates, creating dependency that limits adversarial independence.

The Daisy McGregor transition illustrates the tension clearly. Her DSIT experience made her genuinely valuable to AnthropicOrganizationAnthropicComprehensive reference page on Anthropic covering financials ($380B valuation, $14B ARR at Series G growing to $19B by March 2026), safety research (Constitutional AI, mechanistic interpretability...Quality: 74/100's external affairs function — but her value to Anthropic derived directly from the relationships and institutional knowledge she built while regulating the sector. The one-year lobbying prohibition under UK rules did not prevent her from immediately building on those relationships in non-lobbying capacities.

For AI Safety Research

The flow of researchers from independent safety organizations into frontier labs is particularly significant for the AI safety field's long-term structure. FHIOrganizationFuture of Humanity InstituteThe Future of Humanity Institute (2005-2024) was a pioneering Oxford research center that founded existential risk studies and AI alignment research, growing from 3 to ~50 researchers and receiving...Quality: 51/100's closure dispersed a significant pool of alignment researchers, many of whom moved to commercial lab safety teams. This has implications for research independence: safety work conducted within a lab is subject to that lab's publication decisions, competitive pressures, and governance culture in ways that independent institute work is not.

The MIRIOrganizationMachine Intelligence Research InstituteComprehensive organizational history documenting MIRI's trajectory from pioneering AI safety research (2000-2020) to policy advocacy after acknowledging research failure, with detailed financial da...Quality: 50/100-to-labs flow represents a different variant: researchers trained in a deconfusionist, mathematically rigorous approach to alignment finding that career pathways increasingly run through labs with empirical scaling orientations. Whether this represents productive cross-pollination or absorption of independent safety capacity is contested within the community.

Open PhilanthropyOrganizationOpen PhilanthropyOpen Philanthropy rebranded to Coefficient Giving in November 2025. See the Coefficient Giving page for current information.Quality: 52/100's position as a major funder of both independent safety organizations and, through its EA ecosystem ties, individuals who take lab positions creates a further structural complexity. Funding relationships can shape research agendas before any formal personnel transition occurs.

For Frontier Lab Governance

The government-to-lab direction — exemplified by Paul Nakasone joining OpenAI's board — serves functions beyond regulatory access. Former senior officials bring credibility with national security agencies, potential influence over government AI procurement, and implicit signals about a company's trustworthiness to policymakers. OpenAI's addition of a former NSA director to its safety committee occurred in the context of ongoing scrutiny of its governance following the November 2023 board crisis, suggesting the appointment served legitimacy as well as operational functions.

Criticisms and Concerns

Regulatory Capture Risk

The dominant criticism holds that AI revolving door flows systematically weaken oversight by creating incentives for officials to favor prospective employers, reduce regulator independence through informal access channels, and concentrate policy expertise in a small network that shares industry assumptions. The empirical evidence from adjacent sectors — particularly the patent examiner and HHS appointee studies — supports concern about favoritism effects even when individuals act in subjective good faith.

Critics including the Revolving Door ProjectOrganizationRevolving Door ProjectGovernment accountability project (of the Goodnation Foundation) led by Executive Director Jeff Hauser. Funded by Democracy Fund. Maintains the "Tracking Uses of AI in the Trump Administration" tra... characterize Big Tech's Washington influence strategy as comprehensive and intentional, combining direct lobbying, funding of nominally independent policy groups, and personnel placements to shape the regulatory environment. The framing is contested: industry proponents argue that government genuinely lacks technical capacity and that experienced practitioners filling advisory roles represents necessary expertise transfer rather than capture.

Conflicts of Interest in Specific Cases

The Bill Pulte case — an FHFA director holding disclosed Palantir stock while overseeing Fannie Mae's AI contract with Palantir — illustrates how financial conflicts can coexist with regulatory responsibility even under formal disclosure regimes. Disclosure requirements identify conflicts; they do not resolve them.

The FERC alumni transitions to data center consulting and AI utility advisory roles are structurally significant because AI infrastructure regulation — energy permitting for data centers, grid interconnection policy — will shape the physical capacity for frontier AI development. Former regulators with specialized knowledge of FERC processes are particularly valuable to companies navigating those processes.

Weaknesses in Current Reform Frameworks

Cooling-off periods are the primary regulatory tool, but existing implementations have documented weaknesses. US federal law imposes one-to-two year restrictions on direct lobbying, but "advising" roles that do not involve formal lobbying contacts are generally permitted immediately. UK rules similarly barred Daisy McGregor from lobbying for one year while permitting relationship-building that serves the same access function. Threshold-based restrictions are vulnerable to positional adjustments that preserve substance while evading formal coverage.

Counterarguments

Proponents of personnel mobility argue that government cannot regulate AI effectively without staff who understand it, and that restricting movement would reduce the quality of both government service and subsequent private-sector practice. Some empirical work on revolving doors in financial services finds that the mechanism can improve regulatory quality through effort incentives — analysts seeking future private employment may work harder to demonstrate competence. This "human capital" hypothesis competes with the "regulatory capture" hypothesis in the literature, with evidence supporting both depending on sector and measurement approach.

The LessWrong community, to the extent it has addressed the issue directly, has included voices supportive of temporary private-sector talent rotation into government as a competence-building mechanism — the "revolving door recommendation" framing — without necessarily treating this as equivalent to capture-oriented industry placements.¹⁰

Limitations

Data Availability

No comprehensive, systematic dataset of AI-specific revolving door transitions exists. The named-moves table above is constructed from available research data and is necessarily incomplete. OpenSecrets and similar trackers provide lobbying disclosure data but do not systematically capture advisory relationships, board memberships, or the soft influence channels that may matter more than formal lobbying in AI governance.

Causal Identification

Attributing regulatory outcomes to revolving door dynamics requires identifying what would have happened counterfactually — what a regulator without industry ties would have decided. This counterfactual is unobservable. Studies that find favorable treatment effects (as in the patent examiner research) use quasi-random assignment designs to approximate causal identification, but such designs are not available for AI governance contexts where selection into roles is highly non-random.

Selection Confounds

High-performing individuals are more likely both to move between sectors and to achieve better outcomes in either direction. Observed correlations between personnel transitions and favorable regulatory treatment may reflect underlying quality differences rather than capture effects. The literature documents this limitation without fully resolving it.

Scope Limitations

Most quantitative revolving door research focuses on financial services, pharmaceuticals, and defense procurement — sectors with long regulatory histories and rich disclosure data. AI governance is younger, more internationally distributed, and involves novel institutional structures (safety institutes, model evaluations, compute governance) for which sector-specific evidence does not yet exist.

AI Safety Community Specificity

The safety researcher flow dynamic — between independent organizations, academic labs, and frontier lab safety teams — is particularly poorly documented quantitatively. Informal norms, community social structures, and funding relationships shape where researchers go in ways that are not captured by standard revolving door frameworks developed for government-industry transitions.

Key Uncertainties

• Whether AI safety institutes can maintain genuine independence from frontier labs given structural recruitment dependencies
• Whether the human capital benefits of personnel mobility outweigh capture costs in AI governance specifically
• How international coordination (US/UK/Canada/Japan AISI cross-staffing) interacts with national competitive interests
• Whether the dispersal of FHI researchers into commercial labs has durably shifted the center of gravity of alignment research
• Whether cooling-off period reform or alternative mechanisms (structural separation, enhanced disclosure, public interest representation requirements) would meaningfully improve governance outcomes

Sources