Capitol Hill Climbing
How Big Tech Actually Influences AI Policy
For the next eight weeks, I’ve joined the Cambridge AI Safety Hub’s Alignment Desk! It’s a structured writing programme for researchers on AI safety to stop procrastinating on publishing research and publish significant work. Whilst other members of the team are security whizzes and reformed quant traders, I’m in my usual stomping ground of political economy: the study of money’s power to influence political institutions.
In 2026, few wield this power as deftly as tech firms. Eight of the largest, including Meta, Alphabet, OpenAI, and Microsoft, spent a cool $71 million in federal lobbying in 2025. The cryptocurrency industry’s primary PAC, Fairshake, has already raised $193 million for pro-crypto candidates in the 2026 midterms. Corporations opening their wallets to further their interests is hardly groundbreaking. Yet, AI’s potential capacity to drive existential risk has led firms to fund both sides of the regulatory debate.
Following the success of Fairshake, major tech players on both the pro- and anti-AI regulation sides of the policy debate have spun up their own entities and advocacy organisations. On the anti-regulation side is Leading the Future, a Super PAC which has raised “more than $125m” from donors including venture capitalist Marc Andreessen and OpenAI President Greg Brockman. On the pro-regulation side is Public First Action, a 501(c)(4) whose major donors include Anthropic.
Whilst readers are likely to know the players involved in this fight, this series will hope to set the board in more mechanistic terms. Environmentally rational strategies are limited by an actor’s capabilities; it’s not acute analysis to recommend chess players flip the board. Rather, we want to consider the pieces (levers of power) and how they’re coordinated towards a win condition agreed upon by both opponents (in this case, shaping AI policy bills). Only once we understand each actor’s limitations and comparative advantages can we evaluate how they might coordinate to shape AI policy in the forthcoming Congressional session.
How Can Tech Firms Influence Capitol Hill?
Special interest groups have a smorgasbord of options for influencing policymaking on Capitol Hill. They can pour money into prestigious policy research groups to influence reports in their favour, spin up litigation centres to fight industry battles in court and more. But there are three primary instruments tech firms can use to swing their financial weight in electoral campaigns and legislative bill drafting.
Lobbyists
Lobbyists, you know ‘em, you despise ‘em, you admire ‘em for their capacity to cajole, convince and dance their way through the proverbial raindrops. Tech companies can hire lobbyists to work in-house or on their behalf through major lobbying firms such as Ballard Partners, Akin, Gump et al., and Holland & Knight. Whilst we might imagine lobbyists as whips of Tammany Hall lore offering gifts, dinner and all manner of economic incentives, kickbacks are effectively banned under the Honest Leadership and Open Government Act of 2007, and House and Senate ethics rules.
Instead, the value of a lobbyist lies in their capacity to gather intelligence and convey information where it has the greatest impact. They’ll provide congressional staff with policy briefs, suggest definitional tweaks for technical terms, and even draft legislation that interests the clients they represent and (ostensibly) the electorate of the congressperson or senator. At the same time, they’ll convey back to clients which congresspeople are amenable to their interests, which committees are likely to have friendly members in the coming congressional cycle, who might lead coalitions on a certain piece of policy and against which companies one needs to play defence.
Lobbyists cannot entirely dictate bills, but can narrow the scope of reasonable policy options to bounds acceptable to special interest groups. They set the agenda. These companies have similar goals; they’d like to decrease their liability and regulatory burden, create a more favourable taxation environment and drive down costs to acquire foreign talent, semiconductors etc. This epistemic coherence can crowd out civil society actors. But these firms are one voice among many; one already unpopular with the electorate. So what tactics do tech firms use to ensure their lobbyists set the agenda? Scale and social networks.
Tech firms flood the zone, sending hundreds of lobbyists to the Hill each year at a cost of tens of millions of dollars. As reported by Issue One, Meta had one lobbyist for every eight members of Congress in 2024, ByteDance one for every ten over the same period. Millions more were spent by Microsoft, Amazon and the like. These lobbyists, more often than not, are former Hill staffers or congresspeople with the connections to get meetings and the cultural capital to deliver policies in the language most familiar and agreeable to Hill staffers. Take Meta, of their 89 in-house staffers, a whopping 73% previously worked on Capitol Hill. This phenomenon is commonly referred to as “the revolving door.”
Dark Money Groups
Dark money groups — an ominous name that more than makes up for the obscurity of their vehicle. DMGs are non-profits that fall under section 501(c) of the Internal Revenue Code, a broad church that includes unions, trade associations and civil rights groups. But the gnarly type of DMG we’d like to focus on is 501(c)(4)s, also known as “social welfare” organisations.
These groups are allowed to spend up to 49.9% of their time on political activities, such as lobbying and supporting/opposing candidates. The other 50.1% must be spent on promoting “social welfare,” including issue advocacy and shaping public opinion. I’d argue these acts are also rather political.
The value in DMGs is their anonymity. 501(c)(4)s aren’t required to disclose their donors. That means tech firms can channel millions of dollars into issue advocacy and lobbying without consumer backlash, shareholder scrutiny or reputational risk. 501(c)(4)s can also direct that money onwards to Super PACs, funneling corporate funds into public expenditure committees without disclosure. 501(c)(4)s often exist alongside Super PACs as part of a broader political network focused on a particular issue. For example, Build American AI is the “public education” arm of the accelerationist Super PAC Leading the Future, and the Cedar Innovation Foundation is networked with the pro-crypto PAC Fairshake.
Political Action Committees (PACs)
PACs are the money-spinners, the explicit vehicles for businesses, unions, or other collectives to raise funds from individuals ($5,000 per person per year) to donate to political candidates ($5,000 per candidate per election). They may also give $15,000 annually to a national party committee, whether Republican or Democratic (but never to Ross Perot).
However, the bulk of PAC spending completed by the technology industry is via Super PACs, independent expenditure groups that can raise unlimited capital to campaign for or against candidates, on the condition that donors are disclosed, money is not provided directly to candidates, and they don’t coordinate their actions with candidates. This vehicle arose out of two 2010 court decisions, the landmark Citizens United v. FEC, which found that putting limits on corporations’ political spending violated their First Amendment rights, and Speechnow.org v. FEC, which followed the logic of Citizens United to conclude that “the government has no anti-corruption interest in limiting contributions to an independent expenditure group.”
Super PACs are especially valuable for tech firms on two fronts. As discussed above, given that their goal is setting the policy agenda rather than championing individual candidates, coordination is less required than it would be for, say, a PAC focused on representing interests in a particular congressional district. Moreover, the lack of expenditure restrictions has enabled large capital injections. Fairshake was the 9th-largest overall and the largest industry PAC by independent expenditures in 2023-24. Super PACs also exist to influence state politics, for example, META California, but we’ll focus on federal politics in this particular series.
With that, the board is set. In the next part, we’ll review common strategies of effective lobbying, find evidence of their use in recent congressional cycles and make some initial forecasts of how the game will play out in 2026–28. Understandably, your first instinct may be to flip the board and storm out.