U.S. AI policy is fragmented and in flux: export controls coexist with a permissive domestic stance while worker-protection initiatives falter, increasing legal uncertainty that will shape where and how firms adopt AI and likely strengthen employer bargaining power.
This commentary examines the emerging body of rules, policies and practices governing the development, adoption and use of artificial intelligence (AI) technologies in the United States, and its implications for work and workers. At the federal level, the United States has so far pursued a strategy based on export controls and a relatively permissive regulatory environment with a patchwork of measures to promote responsible AI innovation and use. As the second Trump administration now begins to implement plans to entirely overhaul frameworks adopted under President Biden, however, the situation is more volatile. Major initiatives designed to hold employers accountable and prevent harms to workers, including Biden’s flagship Executive Order, are no longer in place. While some progress can be observed at the state level, many proposals for legislation to strengthen workers’ rights in relation to AI have stalled. A conservative majority in the Supreme Court meanwhile lays the ground for further rulings that could undermine the power of organised labour. Despite these enormous challenges, workers are increasingly regarding AI adoption and use as a site of collective struggle. Alongside jurisdiction case reports on China, Canada, Brazil, India and the EU, the following discussion of the US’s AI regulation, development and governance approaches today is part of the Artificial Intelligence Policy Observatory for the World of Work (AIPOWW) symposium.
Summary
Summary — Gray (2026), "AI governance under the second Trump administration: implications for labour"
Gray, C. 2026. AI governance under the second Trump administration: implications for labour. Global Political Economy, pp.1–12. https://doi.org/10.1332/26352257Y2025D000000052
Main Finding
The rollback of Biden-era AI governance under the second Trump administration—combined with intensified influence from Big Tech oligarchs and a permissive federal AI Action Plan—creates a regulatory environment likely to accelerate private-sector AI adoption while reducing worker protections. This shift increases risks of labour displacement, surveillance, and worsening power asymmetries between employers and workers, even as state-level legislation and union activity produce uneven and partial countervailing pressures.
Key Points
-
Political economy context
- Returning Trump administration rapidly dismantled Biden-era AI governance (notably rescinding EO 14110) and issued an AI Action Plan prioritizing innovation, infrastructure, and international AI diplomacy/security.
- Tech oligarchs and technofuturist/techno‑optimist networks (e.g., Musk, Andreessen, Thiel ties) have heightened influence on federal policymaking.
- The US remains a leading producer of large AI models (Stanford AI Index 2025), but China is narrowing quality/quantity gaps; global competition frames much policy rhetoric.
-
Federal policy changes and instruments
- EO 14110 (Biden): emphasized “safe, secure, trustworthy” AI and contained explicit worker-oriented provisions (DOE’s DOL Principles and Best Practices on AI & worker wellbeing).
- Trump’s 2025 AI Action Plan (and subsequent EOs) focuses on accelerating innovation, building AI infrastructure, and promoting US governance models abroad; it removes or discourages references to non-technical social concerns (e.g., DEI, climate) and ties federal funding incentives to deregulatory aims.
- Executive Orders under Trump do not directly preempt states but threaten withholding federal funds from states with “burdensome” AI rules.
-
Labour- and workplace‑focused developments
- The Department of Labor’s 2024 Principles and Best Practices (produced under EO 14110) recommended worker empowerment, transparency, human oversight, protections for labour rights, upskilling support, and restricted worker data practices—documents now nominally abandoned by the federal executive.
- Trump’s plan emphasizes AI skills training and retraining funds but lacks binding worker protections addressing discrimination, surveillance, or limits on employer uses of AI.
- Employer discretion over deployment of AI is likely to expand; this increases potential for algorithmic management, surveillance, automated decision-making, and weakening of worker protections.
-
Industrial relations and legal context
- US labour law is characterized by voluntaristic, decentralized unionism and strong managerial rights; unionization faces high barriers (Andrias 2019).
- Judicial changes (e.g., Loper Bright v Raimondo weakening Chevron deference) and leadership turnover at agencies (FTC chair changes) reduce administrative capacity and regulatory certainty.
- State-level activity is extensive and heterogeneous (many states passed AI-related laws in 2024), creating a patchwork that can both mitigate and complicate national outcomes.
-
Firms, military, and ethics
- Increased ties between industry and military/security (contracts, partnerships) and relaxation of tech-company self‑restraints (e.g., Google dropping war-use limits) shift firm incentives toward defence and surveillance markets.
- Worker resistance (union statements, strikes, bargaining over AI use) exists but faces constraints; some collective bargaining has produced worker-protective AI clauses in specific firms/sectors.
Data & Methods
- Approach: qualitative policy and legal analysis.
- Sources reviewed: executive orders, federal agency guidance (e.g., NIST AI RMF), Department of Labor principles/best practices, congressional proposals, state legislation, court decisions (notably Loper Bright), technical standards, industry announcements, and secondary literature (academic, media, think tanks).
- No original empirical data collection or quantitative analysis; the paper explicitly states that no empirical human-subject research was conducted and ethics approval was not required.
- Comparative context: jurisdictional case reports on China, Canada, Brazil, India, and the EU are used for contrast in the broader AIPOWW symposium, but the core analysis focuses on US policy and labour implications.
Implications for AI Economics
Policy and market effects — high-level - Faster AI adoption under deregulation should raise aggregate productivity in affected firms/sectors but will likely be highly uneven across workers and regions. - Income and wealth concentration may increase: owners of AI-capital and platform firms (tech oligarchs) gain outsized returns; weaker antitrust/regulatory constraints risk greater market concentration and rent extraction. - Labour market disruption: greater displacement risk for routine and some non-routine tasks; skill-biased technological change increases demand for AI-related skills while lowering wages/conditions for low-skill workers. - Workplace dynamics: expanded algorithmic management and surveillance heighten monopsony power of employers, potential for wage suppression, and higher incidence of non-wage managerial controls.
Research opportunities and empirical strategies - Use state-level policy heterogeneity as quasi-experimental variation: - Difference-in-differences or event-study designs exploiting timing of state AI laws and federal EO rescission/AI Action Plan announcements to estimate effects on firm AI adoption, hiring, wages, and layoffs. - Construct AI-exposure indices at industry or occupation level (text-based measures from job ads, patent/model release links, firm disclosures) to study heterogeneous impacts. - Matched employer–employee administrative or panel survey data to estimate impacts on earnings, hours, job transitions, and re-/up-skilling returns. - Evaluate training programs: randomized evaluations or regression discontinuity designs around eligibility thresholds for federally funded retraining tied to the AI Action Plan. - Measure algorithmic surveillance and discriminatory outcomes via audit studies, complaints/litigation data, and worker surveys; link to firm outcomes to assess productivity vs. welfare trade-offs. - Antitrust and market structure analysis: use firm-level financials and market concentration measures to test whether deregulatory shifts increase markups, decrease labor share, or alter investment patterns.
Policy-relevant economic hypotheses to test - Deregulation-led acceleration of AI adoption increases firm-level productivity but reduces employment and average wages in exposed occupations within 1–3 years. - State AI regulations that include worker-protective measures mitigate negative wage/employment impacts relative to deregulatory states. - Federally promoted retraining initiatives increase re-employment probability but have heterogeneous returns by worker age, education, and local labor-market structure. - Greater AI-driven surveillance correlates with reduced bargaining power and lower wage growth for non-supervisory workers.
Practical implications for economists and policymakers - Need to monitor distributional outcomes (wage inequality, labor share decline) and the incidence of displacement across demographic groups and geographies. - Evaluate the cost-effectiveness of retraining-focused policies versus stronger workplace protections (transparency, bargaining rights, limits on surveillance) for preserving worker welfare. - Recognize legal and institutional constraints (labour law, weakened agency power) when designing interventions; state-level policies and collective bargaining may be the most practicable levers in the near term. - Consider integrating competition policy (antitrust) with labour market regulation to address combined effects of concentration and monopsony.
If you want, I can: (a) extract specific policy timelines and EO citations from the paper into a compact chronology; (b) outline a potential empirical paper design using state law variation to estimate AI’s impact on wages; or (c) produce a one-page brief for policymakers summarizing recommended protections for workers. Which would be most useful?
Assessment
Claims (16)
| Claim | Direction | Outcome | Confidence & Evidence | Details |
|---|---|---|---|---|
| The federal U.S. approach to AI governance combines export controls for key AI hardware/software with a relatively permissive domestic regulatory stance that relies on executive guidance, voluntary standards, and sector-specific measures rather than comprehensive federal worker protections. Governance And Regulation | null_result | regulatory posture / governance instruments at federal level (export controls; presence/absence of comprehensive worker-protection regulation) |
Reading fidelity
high
Study strength
n/a
|
not reported
|
| The incoming second Trump administration is dismantling many Biden-era worker-protection initiatives (notably rescinding or undercutting the Biden Executive Order intended to hold employers accountable for AI impacts). Governance And Regulation | negative | existence and scope of executive-order-based worker-protection initiatives |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| State-level advances in worker-protective AI measures exist but are uneven and many proposed state bills aimed at strengthening workers’ rights related to AI have stalled. Governance And Regulation | null_result | status of state-level legislation regarding AI and worker protections (enacted vs. stalled) |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| A conservative Supreme Court majority increases the risk of rulings that could further constrain organized labor and weaken labor’s power to negotiate AI-related workplace rules. Governance And Regulation | negative | legal constraints on organized labor’s bargaining power (court rulings affecting labor law) |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| Workers are increasingly treating AI adoption as a collective bargaining and political issue, using strikes, bargaining demands, and internal organizing to contest deployments. Worker Satisfaction | positive | worker organizing activity focused on AI (strikes, bargaining demands, internal organizing efforts) |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| Regulatory volatility and fragmentation will shape firms’ AI investment decisions, firms’ workplace practices (surveillance, task allocation), and the distributional consequences of AI for wages, employment and bargaining power. Firm Productivity | mixed | firm AI investment decisions; workplace practices (surveillance, task allocation); distributional labor-market outcomes (wages, employment, bargaining power) |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| Export controls may constrain access to advanced models and hardware, affecting productivity gains unevenly across firms and sectors. Firm Productivity | negative | access to advanced AI models/hardware; sectoral/productivity gains |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| Regulatory uncertainty (rollbacks and a patchwork of rules) can raise compliance and political risk costs, causing some firms to accelerate private governance and self-regulation while causing others to delay investment or relocate activities. Adoption Rate | mixed | firm responses: adoption of private governance/self-regulation; investment timing; geographic relocation |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| Weakening of organized labor and stalled worker-protection legislation raises the probability that AI adoption will increase employer bargaining power, potentially depressing wages and worsening job quality for affected occupations. Wages | negative | employer bargaining power; wages; job quality in affected occupations |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| Absent strong worker voice or mandated impact assessments, AI-driven surveillance, algorithmic management and task reallocation are more likely, increasing risks of deskilling, displacement, and discriminatory outcomes. Job Displacement | negative | incidence of surveillance and algorithmic management; worker outcomes (deskilling, displacement, discrimination) |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| State-level divergence in AI-related regulation will create geographic heterogeneity in adoption costs and labor protections, potentially inducing firm and worker sorting across states and making national inference about AI’s effects more difficult. Adoption Rate | mixed | geographic heterogeneity in adoption costs; firm/worker sorting across states |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| Private governance and firm-level solutions (internal standards, bargaining with unions) may proliferate, but these can entrench firm-specific norms and increase market power asymmetries. Market Structure | negative | prevalence of private governance; firm-specific norms; market power asymmetries |
Reading fidelity
low
Study strength
n/a
|
not reported
|
| Insufficient regulation increases risks of negative externalities (privacy harms, biased hiring/management) that can reduce labor supply attachment or lower human capital investments. Skill Acquisition | negative | privacy harms; biased hiring/management; labor supply attachment; human capital investment |
Reading fidelity
medium
Study strength
n/a
|
not reported
|
| The paper is a policy and legal commentary/synthesis and not an empirical causal study; it does not provide microdata on employment or wage effects but identifies plausible channels and institutional dynamics. Research Productivity | null_result | study type / presence of primary empirical data |
Reading fidelity
high
Study strength
n/a
|
not reported
|
| There is a need for validated administrative and firm-level data on AI adoption, workplace monitoring, and worker outcomes, and for evaluation of policy interventions (mandated impact assessments, transparency requirements, worker representation rules) using randomized or quasi-experimental designs where feasible. Research Productivity | null_result | availability of validated administrative and firm-level AI adoption data; existence of rigorous evaluations of policy interventions |
Reading fidelity
high
Study strength
n/a
|
not reported
|
| Policy recommendations implied include: reinforce worker voice via required worker representation in AI impact assessments and protection of collective bargaining around technology use; mandate disclosure and standardized impact reporting of AI systems used for hiring/monitoring/promotion/termination; and implement targeted sector- or task-specific enforceable regulations. Governance And Regulation | positive | adoption of recommended policy measures (worker representation, disclosure mandates, targeted regulations) |
Reading fidelity
speculative
Study strength
n/a
|
not reported
|