Big Tech Spent $1.1 Billion Shaping AI Law. Here Is the Paper Trail.

Federal lobbying disclosures, FEC records, and peer-reviewed research trace a direct line from nine-figure technology industry spending to a legislative provision that would strip states of AI...

The single most documented fact in the public record of American AI policymaking is this: on the eve of Congress embedding a 10-year moratorium on state AI regulation into federal spending legislation, four of the companies that lobbied for that provision — Microsoft, Amazon, Alphabet, and Meta — were collectively spending tens of millions of dollars per quarter on federal influence operations, against a backdrop of at least $1.1 billion in total Big Tech political spending during the 2024–2025 cycle alone. That figure comes not from an advocacy group's estimate but from Public Citizen's 2025 analysis of FEC filings, corporate disclosure reports, and documented inauguration contributions. The spending preceded the legislative outcome. The legislative outcome benefited the spenders. The public record makes both facts simultaneously and verifiably true.

The architecture of that spending is worth examining in granular detail, because the aggregate number obscures the velocity of individual actors. According to Issue One's 2026 lobbying tracker, which draws from Senate Office of Public Records LD-2 filings, Anthropic — a company that publicly brands itself as a safety-focused AI developer and that has received substantial U.S. government AI contract interest — increased its federal lobbying expenditure by 333 percent in a single year, from $360,000 in Q1 2025 to $1.6 million in Q1 2026, or approximately $17,000 per day. Microsoft filed Q1 2026 lobbying expenditures of $2.6 million, roughly $29,000 per day. Nvidia reported $1.3 million, a 37 percent year-over-year increase. These figures cover only three companies and only one quarter. The Q1 2026 LD-2 filings for Alphabet, Meta, Amazon, and OpenAI — all historically higher-spending entities — had not been fully itemized in available source materials at the time of this analysis, meaning the figures cited here represent a floor, not a ceiling.

The growth trajectory of AI-specific lobbying is itself a primary data point. According to OpenSecrets data cited by Nemko Digital Intelligence, six organizations lobbied the federal government on AI-related issues in 2016. By the first quarter of 2023 alone, that number had reached 123, with collective lobbying spend across all issues totaling approximately $94 million in that single quarter, per OpenSecrets' May 2023 analysis. Oracle, to select one company from the public record, reported $3.1 million lobbied specifically on AI and machine learning policy in Q1 2023. By 2025, more than 450 organizations were engaged in AI lobbying — a 7,567 percent increase from the 2016 baseline. This expansion tracks with precision against two independent variables: the commercial deployment of large language models beginning in late 2022, and the emergence of AI-specific legislative proposals at the state and federal level. The industry mobilized in direct proportion to the regulatory threat it perceived.

The 10-year state preemption provision embedded in President Trump's spending legislation is the clearest documented case of lobbying expenditure converting into legislative text. The provision, referenced in a House Judiciary Committee document filed with Congress (HHRG-119-JU05-20251216-SD014-U14), would prohibit all U.S. states from enacting or enforcing their own AI or social media algorithm regulations for a decade. The coercion mechanism is structural: states that declined to cede that authority would forfeit federal broadband funding, making non-compliance financially untenable for most state governments. Microsoft, Amazon, Alphabet, and Meta are named in Issue One's reporting as corporate advocates for the provision. Issue One's characterization — that the industry was 'trying to cash in on years of lobbying' — is analytically consistent with the spending timeline, though the causal relationship between lobbying expenditure and specific legislative language cannot be established from financial disclosures alone. What can be established is the sequence: the money came first, the provision came second, and the named beneficiaries of the provision are the named sources of the money.

The personnel dimension of this influence operation is quantified in a peer-reviewed paper published at arxiv.org under the title 'Big AI's Regulatory Capture: Mapping Industry Interference and Government Complicity.' Analyzing two independent datasets of high-profile U.S. AI policy cases, the researchers found revolving door presence — meaning personnel movement between government regulatory roles and industry positions — in 24 percent of the cases in their primary dataset. One in four major AI policy interactions, by that measure, involves someone who has worked on both sides of the regulatory relationship. The RAND Corporation, in a December 2024 research brief drawing on interviews with 15 of 17 experts who cited industry influence concerns, assessed that the AI industry had gained 'extensive influence' in general-purpose AI regulation conversations, and that the structural preconditions for full regulatory capture — defined as industry priorities systematically overriding public interest through entrenched mechanisms, not incidental access — were materially in place, even if the threshold had not yet been formally crossed. That assessment predates the 2025 preemption provision victory.

The European Union record adds a parallel data stream. The Corporate Europe Observatory, in research by Schyns (2023) cited in the arxiv paper, documented that technology companies had, through years of direct pressure, covert industry groups, and the deployment of industry-funded experts presenting as independent voices, achieved measurable outcomes in the EU AI Act: reduced safety obligations, sidelined human rights and anti-discrimination provisions, and regulatory carveouts for specific AI products. The 'tech-funded experts' formulation describes a specific mechanism — researchers or witnesses whose independence is structurally compromised by financial relationships with the industry they are nominally evaluating — that operates outside the formal lobbying disclosure system and is therefore largely invisible to standard financial analysis. The full Corporate Europe Observatory report names specific individuals and organizations; the excerpted source materials available to this analysis do not.

Three gaps in the public record define what remains hidden and what instruments would reveal it. First, the 10-year preemption provision's final legislative status — whether it survived Senate conference or was modified — requires direct monitoring of the Senate Legislative Information System and the enrolled bill text. Second, the arxiv paper identifying 10 named high-profile revolving door cases truncates before listing those individuals; the full paper at arxiv.org/html/2605.06806v1 contains names, prior government roles, and subsequent industry positions that this analysis cannot independently confirm from available excerpts. Third, and most structurally significant: U.S. federal lobbying disclosure forms do not require companies to allocate expenditures by specific issue, meaning the fraction of any company's quarterly lobbying spend directed specifically at AI policy versus antitrust, tax, or trade issues cannot be determined from LD-2 filings alone. That transparency gap — a feature of the disclosure system, not a bug in this analysis — is the single reform that would most directly illuminate the money-to-policy connection that the existing record only partially documents. A mandatory issue-level cost allocation requirement in LD-2 filings, applied uniformly to all filers, would close it.