OpenAI, Anthropic and NVIDIA spent a combined $7.5 million lobbying the federal government in the third quarter of 2026, about 23% more than the second quarter and 61% more than the same period last year, according to an investigation by the Project On Government Oversight based on newly filed disclosure reports. The Verge first reported that the quarter was a spending record for each of the three companies individually: NVIDIA at $2.2 million, Anthropic at $2.1 million and OpenAI at $1.5 million. Since the beginning of 2024 the three companies have spent about $37 million on federal lobbying — roughly half of that in 2026 alone.
The Growth Trajectories Behind The Records
Each company’s climb looks sharp against its own history. NVIDIA’s federal lobbying spending has risen nearly ninefold, from $640,000 in 2024 to $6 million in 2025, and has already reached $6.04 million in 2026 with a quarter remaining — a 53.5% jump from Q2 to Q3. This was NVIDIA’s first quarter above $2 million. Anthropic’s annual lobbying spend has risen nearly eightfold over three years: $1.04 million in 2024, $3.81 million in 2025 and $8.11 million so far in 2026, with one outside firm’s late disclosure still adding to the total. OpenAI spent $2.45 million in 2024, $4.14 million in 2025 and $4.77 million so far this year, putting it on pace to double its 2024 total.
The sums remain small next to Washington’s established players — Microsoft has spent at least $2 million a quarter since 2021 and Google regularly spends over $3 million — but the velocity is the story. None of the three AI companies was a serious Washington presence five years ago, and all three have built their lobbying operations almost entirely since 2024.
What They Are Pushing On
The disclosure filings list the policy fights driving the spend. NVIDIA’s lobbyists have focused heavily on semiconductor trade policy and export controls, targeting Congress, multiple federal agencies and White House offices including the National Security Council and the National Economic Council. Anthropic’s filings cover AI-related provisions in the Kids Online Safety Act, export control provisions in the AI Overwatch Act, the executive order renaming AI “super intelligence” and the Ratepayer Protection Act, which would put data-center energy costs on developers. OpenAI engaged on cybersecurity evaluations for frontier models, AI safety oversight frameworks, copyright and antitrust protections for information sharing, contacting Congress, the White House, the Small Business Administration and the Commerce and State Departments.
The context is a Washington turning openly adversarial toward parts of the industry. Federal lawmakers have introduced at least 99 AI-related bills in the current Congress, the Federal Trade Commission has sent civil subpoenas to Anthropic and OpenAI seeking information about security incidents and privacy policies, and the companies have hired firms with ties to the current administration — OpenAI brought on Miller Strategies after the 2024 election, while Anthropic hired Ballard Partners in April after its dispute with the Pentagon reached a boiling point.
Why It Matters
The record quarter captures the industry’s bet that policy fights will be decided in Washington, not in product launches. Each company is spending to shape a different risk: NVIDIA is defending chip exports, Anthropic is defending its safety-first positioning against Pentagon blacklisting pressure, and OpenAI is managing fallout from incidents like the Hugging Face breach while fighting copyright and safety oversight rules. The POGO analysis notes that roughly half of the $37 million spent since 2024 arrived in 2026 alone — the money is front-loaded because the legislative window is. Bills on export controls, data-center energy costs and child safety could each reorder the companies’ cost structures, and the lobbyists are arriving before the laws do.
Background
The AI industry’s Washington presence has grown alongside its regulatory exposure. Anthropic’s own draft legislation — which would let Washington block a model only above $1 billion in damages or 50 deaths — was revealed in October after the company pitched it to lawmakers, and federal agencies from NIST to CISA have moved to set rules for AI agents acting on their own. The POGO findings add the money layer: as the EU AI Act and similar frameworks take shape abroad, the domestic fight is increasingly being fought through federal filings, one lobbying disclosure at a time.
Source: Project On Government Oversight
0 Comments