The fight
Every dollar spent lobbying on artificial intelligence in Washington is spent on one question: who writes the rules — the federal government, or the states.
That is the fight. Everything else — copyright, deepfakes, data centers, export controls, child safety — runs downstream of it. The companies building frontier AI models have decided that fifty separate state regimes is the one outcome they cannot operate inside. A smaller set of AI firms, plus advocacy groups and state legislators, have decided that a single light-touch federal standard is the outcome they cannot accept. Both sides are now spending at a scale technology policy has never seen.
Here is who is paying — and what they are buying.
The money
2025 was the first year tech and AI companies spent more than $100 million on federal lobbying, according to Bloomberg figures. Meta alone reported $26.29 million — more than any company in any industry. Amazon spent $17.89 million, Alphabet $13.10 million, Microsoft $9.36 million. Nvidia spent $4.9 million — roughly seven times its 2024 figure.
The firms taking the work did well. Lobbying shops collected close to $92 million from AI-related issues in the first three quarters of 2025, and roughly $130 million for the full year, per Bloomberg Government analysis.
The pace then accelerated. In the first quarter of 2026, eleven top tech companies spent $20 million on federal lobbying — about $226,000 a day — according to the reform group Issue One. Meta led again at $7.1 million, with Amazon at $4.4 million and Google at $2.9 million.
The newer development: the frontier labs are no longer bystanders. Anthropic spent $1.6 million in Q1 2026, its largest quarter ever — up from $360,000 a year earlier, a jump of nearly 494%. OpenAI spent roughly $1 million, also a record, close to double its year-earlier figure. OpenAI opened its first Washington office; Anthropic is expanding its D.C. footprint.
The labor market moved with the money. A New York Times analysis by Cecilia Kang found that roughly one in four of the country's ~13,000 registered federal lobbyists now work on AI at least once a year — up from 11% in 2023.
What they want — three camps
The map is not company-by-company. It is position-by-position. There are three.
Camp one: preemption
The largest bloc — Meta, Google, Amazon, OpenAI, Andreessen Horowitz, Nvidia. The goal is one federal AI framework that overrides state law. The stated reason: a fifty-state patchwork makes compliance impossible and cedes ground to China. The reason critics name: a single federal standard, written light, is far easier to live with than fifty state experiments — some of which, on liability and safety disclosure, have teeth.
This camp nearly won in 2025. Senator Ted Cruz tried to attach a state-AI-preemption provision to the GOP budget package; bipartisan resistance stripped it out. It returned as executive action — a December 11, 2025 White House executive order pushing federal AI standards and aimed at limiting state regulation. And it survives as legislation: Representative Baumgartner's American AI Leadership and Uniformity Act would write preemption into statute, and Cruz's SANDBOX Act would create a federal regulatory sandbox for AI.
Camp two: guardrails
Smaller, louder — and, notably, it includes an AI company. Anthropic self-reported a $20 million donation to Public First Action, a bipartisan network whose priorities invert the preemption camp's: preserve state authority over AI, protect whistleblowers who expose dangerous AI activity, and require advanced-AI developers to disclose what they are building before release. Brad Carson, a former Democratic congressman, runs the affiliated Public First, which has pledged roughly $50 million for the midterms.
This is the single most important feature of the AI lobbying map: it is not industry versus government. It is industry versus industry. Anthropic and OpenAI — the two leading frontier labs — are on opposite sides. In Illinois, they are backing competing bills on whether frontier developers can be held liable for catastrophic harm; OpenAI supports a version that would not impose liability for an AI system that causes 100 or more deaths or more than $1 billion in property damage.
Camp three: the infrastructure and vertical buyers
The quiet majority. Cloud providers, chip designers, data-center operators, and the "AI for X" startups in healthcare, defense, education, and finance — many retaining Washington counsel for the first time. They are not fighting over preemption. They want procurement access, export rules, energy and permitting for data centers, and favorable treatment inside their own verticals. Nvidia's lobbying — up sevenfold — tracked a single outcome: the reversal of the ban on selling its most advanced chips to China. This camp is why AI lobbying spend keeps compounding even when a given bill dies. There is no one law that ends it.
The super PAC war — built on the crypto template
Lobbying is the inside game. The outside game is the super PACs — and it is explicitly modeled on crypto.
In 2024, the crypto industry spent more than $130 million through super PACs, chiefly Fairshake, and remade the Senate's posture toward digital assets. The AI industry watched, and copied the playbook.
Leading the Future — backed by Andreessen Horowitz, OpenAI president Greg Brockman ($12.5 million), Palantir co-founder Joe Lonsdale, SV Angel's Ron Conway, and Perplexity — has raised roughly $140 million for the 2026 cycle, entering the year with about $70 million cash on hand, per FEC filings.
Its first and clearest target: Alex Bores, a New York Democrat and former Palantir engineer running for retiring Representative Jerry Nadler's seat. Bores co-sponsored New York's RAISE Act, a state AI safety law. Leading the Future ran ads against him before he had cleared a primary.
A second pro-industry network, Innovation Council Action — championed by White House AI adviser David Sacks — plans to spend upward of $100 million and has built a scorecard ranking lawmakers on alignment with the administration's AI agenda. Meta is backing a separate, state-focused super PAC reported in the $65 million range.
On the other side stands Public First — roughly $50 million, pro-regulation, bipartisan. The asymmetry the pro-regulation side is counting on is public opinion: polling on AI runs skeptical. Carson's wager is that sentiment closes the money gap.
The bills
The track that moves: deepfakes and likeness
The TAKE IT DOWN Act, banning nonconsensual intimate imagery including AI-generated deepfakes, was signed in May 2025. The Senate passed the DEFIANCE Act in January 2026, letting deepfake victims sue for up to $250,000. The NO FAKES Act covers voice, image, and election deepfakes. Bipartisan, low-controversy, victim-focused — it passes.
The track that is stuck: preemption itself
The GUARDRAILS Acts and a States' Right to Regulate AI Act would void the December executive order outright. The American AI Leadership and Uniformity Act and the SANDBOX Act would lock preemption in. Neither side has the votes. Which is precisely why both sides are spending more than $100 million on the midterms — the votes are what the money is trying to buy.
The states are not waiting
While Washington is gridlocked, the states are writing the actual rules.
In California, AI and crypto firms spent $39 million influencing state politics in 2025, per a CalMatters analysis. Meta spent at least $4.6 million lobbying California officials — its largest figure since it began advocating in Sacramento in 2010. Meta and Google each put $5 million into a single election committee. In New York, the RAISE Act. In Illinois, Anthropic and OpenAI on opposite sides of a frontier-AI liability bill.
The preemption camp is spending federally because it is losing locally. Federal gridlock plus a deregulation-minded White House means the AI rules with real force today are being written in Sacramento, Albany, and Springfield. A federal preemption statute is the off-switch for all of it.
What it means
For anyone in communications, public affairs, or policy, the AI regulation fight is not a 2027 story. It is being decided now — in primaries, in state committee rooms, and in quarterly disclosure filings. And the organizations shaping the outcome are not only the frontier labs. They are trade associations, vertical startups, and advocacy coalitions, most of which have been in this fight for under two years.
The venue where that fight gets explained — to voters, to staffers, to journalists, and increasingly to the AI engines now summarizing "who is lobbying on AI" — is itself contested. The narrative is part of the lobbying.
One question decides it: Washington, or the states. Watch the midterms. Watch the filings. The money is already telling you who expects to win.
