Redistributed

OpenAI spent $260,000 on federal lobbying in 2023. By 2025, that figure had climbed to nearly $3 million. The company increased its lobbyist count from 3 to 18 in a single year. This trajectory tells you everything you need to know about where the AI industry believes the real decisions are being made.

The question isn't whether AI will be regulated. It's who writes the rules. And right now, the companies building AI systems are spending hundreds of millions of dollars to ensure the answer is: largely themselves.

The Scale: Money Talks

The numbers are staggering. In 2025, the top tech and AI companies spent more than $100 million to influence federal policy:

Company	2025 Lobbying Spend
Meta	$26.29 million
Amazon	$17.89 million
Alphabet (Google)	$13.10 million
Microsoft	$9.36 million
OpenAI	~$3 million
Anthropic	~$1.5 million

The growth rate matters as much as the absolute numbers. OpenAI's lobbying spend increased nearly seven-fold from 2023 to 2024, then nearly doubled again in 2025. Anthropic more than doubled its spend from 2023 to 2024. The number of organizations lobbying on AI jumped from 158 in 2022 to over 648 in 2024 — a 141% increase in two years.

This isn't companies responding to regulation. It's companies shaping regulation before it arrives. To be fair, some of this engagement reflects genuine industry expertise that policymakers need. The problem is not that AI companies participate in policy — it is the scale of the asymmetry between their participation and everyone else's.

What They're Lobbying For

AI companies publicly call for regulation. In private, they work to ensure that regulation serves their interests. The pattern, documented repeatedly by researchers and watchdog groups, involves three core strategies:

1. Liability Shields

The big ask: extend something like Section 230's immunity to AI outputs. Section 230 shields platforms from liability for content users post. AI companies want similar protection for content their models generate.

The argument: holding companies liable for AI outputs would "harm innovation" and damage "competitiveness." Senators Josh Hawley and Richard Blumenthal introduced the No Section 230 Immunity for AI Act to prevent this extension. It went nowhere. As one legal analysis noted, AI companies "should not assume Section 230 immunizes them from claims like defamation, privacy violations, or false information harms" — but they're lobbying to make that assumption reality.

2. Self-Regulation Over Mandatory Rules

The Frontier Model Forum — founded by OpenAI, Anthropic, Google, and Microsoft — claims it won't engage in lobbying. But the forum's existence is itself a lobbying argument: we can govern ourselves, so you don't need to govern us.

The voluntary commitments these companies made to the White House in 2023 are the template. They promised to watermark AI content, share safety information, and test models for dangerous capabilities. None of these commitments are legally binding. A year later, MIT Technology Review found implementation was "uneven at best."

Self-regulation serves incumbents twice: it avoids mandatory rules, and it creates compliance theater that smaller competitors can't match. When you define "responsible AI" as requiring extensive safety teams, red-teaming infrastructure, and formal governance processes, you've described resources only the largest companies possess.

3. Federal Preemption of State Laws

The preemption battle is where the real action is. California's SB 1047 would have held AI developers liable if their models were modified and used for harm. The industry mobilized against it: OpenAI, Google, Meta, and Andreessen Horowitz all lobbied for its defeat. Governor Newsom vetoed it.

The industry argument: companies can't comply with fifty different state regulatory regimes. The real concern: state legislators are moving faster and more aggressively than federal regulators. California, Colorado, and New York have all advanced AI oversight bills that go beyond anything proposed federally.

Big Tech's combined political spending exceeded $1.1 billion in 2024-2025, much of it aimed at preventing state-level regulation. The strategic goal is federal preemption — a single, weaker federal framework that wipes state laws off the books. OpenAI has explicitly lobbied for federal regulation while opposing state regulation. The contradiction resolves when you understand that "federal regulation" means regulation they can influence, while state regulation means fifty different fights they might lose.

The Revolving Door

Money is only part of the influence machine. Personnel flows matter as much.

Former Senator Laphonza Butler (D-Calif.) left Congress and became a consultant to OpenAI. Former Representative Jeff Miller (R-Fla.) represents Palantir as a registered lobbyist. Former UK Deputy Prime Minister Nick Clegg spent six years as Meta's global policy chief. Former Prime Minister Rishi Sunak took advisory roles with Microsoft and Anthropic after leaving office.

Mistral AI opened a Brussels lobbying office headed by Cédric O, France's former secretary of state for digital transition — a man known to have the ear of President Macron.

The revolving door spins both directions. Former tech executives join government; former regulators join tech. Each rotation transfers knowledge, relationships, and assumptions about what's possible. When the people writing AI rules spent their careers building AI companies, the rules tend to reflect industry perspectives.

As Lee Fang documented, "Members of Congress increasingly take jobs as AI lobbyists." This isn't corruption in the legal sense. It's something more structural: a system where the boundaries between regulator and regulated blur until they're hard to see at all.

Industry Associations: "Responsible AI" as Strategy

The AI industry has built a dense network of associations, initiatives, and partnerships that present industry preferences as public-interest positions:

The Frontier Model Forum — Founded by OpenAI, Anthropic, Google, and Microsoft. Presents itself as focused on "safety research," not lobbying. But the forum's very existence argues that frontier AI is best governed by frontier AI companies.

Partnership on AI — A multi-stakeholder initiative that includes civil society groups alongside companies. The intent is genuine — bringing diverse perspectives together — but critics note the power imbalance: corporate members have resources to staff working groups and shape agendas that nonprofits can't match.

The AI Alliance — A coalition that explicitly opposed California's SB 1047, framing liability requirements as threats to open-source AI. The alliance presents industry opposition as protecting the open ecosystem.

Information Technology Industry Council (ITI) — A trade association representing the largest tech companies. Actively lobbies on AI policy, data privacy, and platform regulation.

These organizations perform a useful function for their members: they translate company interests into public-interest language. "Protecting innovation," "ensuring competitiveness," "enabling the open AI ecosystem" — these phrases appear in both trade association statements and legislative testimony. They become the vocabulary of the debate.

The EU Fight: A Case Study

The EU AI Act negotiations revealed how industry lobbying works in practice.

Corporate Europe Observatory documented the campaign: Big Tech companies held more than one lobby meeting per day with Commission officials during the AI Act negotiations. Fully 86% of high-level Commission meetings on AI were with industry representatives.

The outcome was measurable. The final Act largely exempted "general purpose AI" from regulation, requiring only transparency obligations. Safeguards like fundamental rights impact assessments were removed. Foundation model providers escaped the strictest requirements.

The lobbying employed what researchers called "covert tactics" — including funding groups that claimed to represent startups while advocating Big Tech positions. When Mistral, a French AI company, advocated for weaker rules, it bolstered the narrative that regulation would harm European competitiveness. That Mistral is backed by significant venture capital and positioned to benefit from weak oversight went largely unmentioned.

The EU example shows the playbook: public calls for "smart regulation" alongside private pressure for maximum exemptions. Brussels, where Big Tech spending on lobbying hit record highs, is where this played out most visibly. But the same dynamics shape every jurisdiction where AI rules are written.

Safety as Regulatory Capture

Here's the uncomfortable part: some AI safety advocacy serves industry interests.

A Springer-published study on "AI safety and regulatory capture" documents the mechanism. When large companies emphasize catastrophic AI risks requiring sophisticated governance, they're also emphasizing problems only large companies have resources to address. Safety requirements that demand extensive red-teaming, alignment research, and formal evaluation frameworks favor incumbents over startups.

The researchers found that "captured AI-safety regulations are likely to create a variety of harms and injustices" by concentrating the industry and reducing competition. More provocatively, critics argue that some companies are "exaggerating safety risks to justify regulation that benefits established players."

This doesn't mean AI safety concerns are fabricated. The risks are real. But the framing of those risks, and the solutions proposed, often align suspiciously well with incumbent interests:

Emphasize catastrophic risks that require massive investment to assess
Propose governance mechanisms that only well-resourced companies can implement
Frame technical safety as requiring proprietary research that shouldn't be open-sourced
Position voluntary commitments as adequate alternatives to legal requirements

The result: a "sophisticated regulatory capture strategy" dressed in safety language. "Safety washing" — performing safety practices for public relations value without substantive protections — becomes indistinguishable from genuine safety work.

The Asymmetry: Who's Not in the Room

The AFL-CIO joined AI lobbying in 2023. So did the NAACP. Mozilla Foundation spent $120,000 on AI advocacy. The Center for AI Safety Action Fund spent $80,000.

Compare that to Meta's $26 million or Amazon's $18 million.

Civil society is outspent by Big Tech by ratios of ten to one or higher. The organizations representing workers, communities, and the public interest operate on budgets smaller than what major tech companies spend on a single lobbying priority.

This asymmetry matters because it shapes what questions get asked. When AI policy discussions focus on "innovation" and "competitiveness," that reflects who's in the room. Questions about worker displacement, algorithmic discrimination, and community impact get less attention — not because they're less important, but because fewer lobbyists are paid to raise them.

Affected communities have the least lobbying capacity. The workers whose jobs are automated, the people subjected to algorithmic decision-making, the communities hosting data centers — they're not at the table. Their interests are represented, if at all, by under-resourced nonprofits competing against the best-funded lobbying operations in the world.

What Redistribution Looks Like

The lobbying asymmetry is itself a redistribution mechanism. When companies shape the rules that govern them:

Value flows to shareholders rather than workers or communities
Liability shifts to users rather than developers
Oversight stays voluntary rather than mandatory
State-level protections get preempted rather than strengthened
Technical standards get set by industry rather than democratic institutions

The alternative isn't eliminating industry input — it's balancing it. Public interest advocates need resources to match corporate lobbying. Regulatory agencies need expertise independent of industry. Affected communities need seats at the table.

Some concrete possibilities:

Public funding for public interest lobbying — The EU requires transparency about lobbying meetings. The U.S. could go further: publicly fund civil society participation in regulatory processes.

Cooling-off periods — Longer gaps between government service and industry employment would slow the revolving door.

Regulatory capacity — Agencies need in-house AI expertise, not dependence on seconded industry employees.

Mandatory disclosure — Real-time disclosure of lobbying contacts, not quarterly summaries.

Affected-community representation — Formal roles for workers, impacted communities, and civil rights organizations in AI governance bodies.

None of this stops lobbying. But it changes who gets heard.

The Stakes

AI systems are becoming infrastructure — embedded in employment, credit, healthcare, education, and justice. The rules written now will shape that infrastructure for decades.

The question of who writes those rules is, ultimately, a question of who AI serves. An industry that spends $100 million per year shaping regulation expects that investment to pay returns. Those returns come from favorable rules: limited liability, self-governance, preemption of stricter standards.

Understanding lobbying as a redistribution mechanism clarifies what's at stake. When companies influence their own oversight, value flows toward them. When affected communities are excluded from rule-making, their interests get discounted.

The influence machine isn't secret. The disclosure data is public. The revolving door is documented. The spending is reported. But the asymmetry persists because transforming money into policy influence is, quite simply, how the system works.

Changing that requires understanding it clearly — and then organizing to counterbalance it.

Sources

OpenSecrets. "Lobbying on AI reaches new heights in 2024." June 2024. OpenSecrets
OpenSecrets. "The AI-fication of K Street." February 2026. OpenSecrets
MIT Technology Review. "OpenAI has upped its lobbying efforts nearly seven-fold." January 2025. MIT Technology Review
TechCrunch. "AI companies upped their federal lobbying spend in 2024 amid regulatory uncertainty." January 2025. TechCrunch
Axios. "How AI swallowed tech lobbying in 2025." January 2026. Axios
Corporate Europe Observatory. "Big Tech lobbying is derailing the AI Act." November 2023. CEO
Corporate Europe Observatory. "Don't let corporate lobbying further water down the AI Act." March 2024. CEO
Fang, Lee. "Members of Congress Increasingly Take Jobs as AI Lobbyists." 2024. Lee Fang
TIME. "Tech Giants Are Vastly Outspending Newcomers on AI Lobbying." May 2024. TIME
Carnegie Endowment for International Peace. "All Eyes on Sacramento: SB 1047 and the AI Safety Debate." September 2024. Carnegie
MIT Technology Review. "AI companies promised to self-regulate one year ago. What's changed?" July 2024. MIT Technology Review
Springer. "AI safety and regulatory capture." AI & Society, 2025. Springer
Congress.gov. "Section 230 Immunity and Generative Artificial Intelligence." Congress.gov
Harvard Law Review. "Beyond Section 230: Principles for AI Governance." 2025. Harvard Law Review

General

The Influence Machine