While politicians debate the safety of artificial intelligence, tech donors are already buying themselves the Congress they need — carefully, legally, and with all the right rhetoric about competing with China
AI Super PACs in the US no longer look like some exotic side feature of major politics. They have already become a full-fledged tool for redistributing power — not through debates about the future of technology, but through a simple and old Washington formula: money enters primaries before voters even have time to understand what exactly is being purchased around them. Formally, the language is about “innovation,” “national competitiveness,” and a “uniform federal approach.” In practice, it is about who will get to write the rules for an industry whose products have long been growing faster than the state’s ability to restrain them.
What is most revealing in this story is not even the amount of money, although that is eloquent enough. What is more interesting is that within the artificial intelligence industry itself, a civil war has already broken out. One faction wants the softest regulation possible and would like to strip states of the power to impose tougher restrictions. The other agrees to some safety measures, but also without excessive zeal and with a very cautious attitude toward real limits. In other words, this is not a fight between people who want oversight and people who want chaos. It is a fight between those who want to regulate the market gently and those who would prefer the market regulate itself — ideally to the applause of Congress.
New York as a testing ground: who exactly the industry wants out of the way
The clearest showcase for this battle has become New York’s 12th District, where in the Democratic primary the industry has already spent more than $2.4 million against state Assemblymember Alex Bores. The sum itself says a great deal. When that kind of money enters a local intraparty race, this is not about affection for one candidate or a random interest in one district. It is about a specific person becoming a political problem for powerful sponsors.
The reason is simple: Bores was one of the authors of one of the country’s first major AI safety bills and made no secret of the fact that he opposed industry efforts to water such rules down. In other words, he did exactly what politicians usually promise voters in theory: he tried to place limits on an industry that is becoming too powerful too quickly. After that, the industry, predictably enough, decided to explain to him the cost of excessive enthusiasm.
The ad campaign against Bores has been built with surgical precision. He was declared the first target to be eliminated, and his role in helping pass the RAISE Act was turned into something close to political contraband. The emphasis on his past work at Palantir is especially elegant. The irony here is fairly dense: one of the donors to the Republican Super PAC working for softer AI regulation is Palantir co-founder Joe Lonsdale. In back rooms, this sort of arrangement is usually called not a contradiction but industry harmony.
Money speaks louder than ideology
The deregulatory side is represented by the Leading the Future network and its partisan arms — the Democratic Think Big and the Republican American Mission. According to Federal Election Commission filings, they raised a little over $50 million last year, although the network itself claims it raised more than $125 million and entered 2026 with roughly $70 million in cash on hand. Even leaving aside the gap between official filings and self-promotional presentations, the scale is already obvious.
The donor list reads like a crash course in the new digital aristocracy. OpenAI president Greg Brockman and his wife Anna Brockman gave $12.5 million each. Marc Andreessen and Ben Horowitz of a16z gave another $12.5 million each. Other notable names include Joe Lonsdale, Ron Conway, and Perplexity AI. The official justification for all this generosity sounds noble enough: America must not lose to China, and it needs a single national regulatory regime. But follow the logic of the messaging and a curious structure appears: in order to protect democratic leadership in technology, one must first spend tens of millions making lawmakers more convenient.
That is exactly what the network’s strategy is built around. Its preferred candidates are meant to support a single federal system for AI regulation that would effectively bar states from adopting tougher rules. The phrasing sounds like a call for uniformity. In essence, it is a federal safeguard against unpleasant surprises from below — to make sure that no state decides safety, transparency, and limits on developers matter a little more than the industry would like.
Moderate, convenient, predictable
Think Big does not merely attack inconvenient candidates. It also selects and promotes Democrats who look sufficiently moderate, business-friendly, and unlikely to moralize too much about technology. Among its recent favorites are Josh Gottheimer, Sam Liccardo, and Yvette Clarke. The last of these, notably, has proposed new rules for AI systems that affect housing, employment, and credit. But there is no real paradox here. The industry does not need to support only those who reject regulation altogether. It only needs to support those who talk about regulation within safe and manageable boundaries.
That is one of the defining tricks of this moment. Big money in politics rarely arrives carrying a sign that says “we oppose rules.” It is much more effective to say, “we support reasonable standards, but without overreach.” From there, the real work begins: making sure the word “reasonable” lines up neatly with the interests of the industry’s biggest players.
The other faction has cleaner hands, but not emptier pockets
The opposing camp within the industry looks more responsible, but only at first glance. Its main engine is Anthropic, which donated $20 million in February to Public First Action. Altogether, that organization has raised at least $50 million and operates through two Super PACs — Democratic Jobs and Democracy PAC and Republican Defending Our Values PAC.
Publicly, this faction says it supports safety standards, transparency rules, and the right of states to impose stricter standards of their own. That certainly sounds better. Almost like the voice of technological conscience. But the official version again looks less than convincing when one recalls Anthropic’s recent behavior. In 2024, the company intervened in California’s AI safety bill, SB 1047, and opposed an approach that would place limits on risk before harm occurs. Instead, it pushed for a softer liability regime based on outcomes rather than preventive restrictions. The basic idea was simple enough: let us not interfere too much with development, and we can talk about risks later, preferably in a flexible format.
In the end, the tougher California bill was vetoed, and a weaker alternative took its place, requiring mostly the publication of safety frameworks and transparency reports. In other words, even among the camp that supposedly favors caution, caution still has clearly market-friendly limits. No one wants to strangle the industry. Everyone just wants to adjust its tie slightly.
So far, the deregulatory camp is barely losing
At the moment, the deregulatory movement led by Leading the Future has suffered only one loss in House primaries across the nine districts where it has intervened. That is no longer an experiment or a lucky streak. It is a functioning political investment model with a very respectable return.
They helped centrist Democrat Melissa Bean win the primary in Illinois’s 8th District by pouring more than $1 million in outside spending into her race through Think Big. Bean, a former Blue Dog Democrat who later became vice chair of JPMorgan Chase, defeated progressive candidate Junaid Ahmed, who had the backing of Bernie Sanders and Elizabeth Warren, by roughly five points. The arithmetic here is ruthless: when a million-dollar outside campaign lands in a race, the idea that this is a pure clash of ideas begins to sound a little naïve.
Republican primaries in Texas have also proved fertile ground. Chris Gober in the 10th District is particularly illustrative. A former chief counsel to Elon Musk’s America PAC, he ran on a platform of increasing AI investment to outcompete China. American Mission spent $747,000 backing him. The ads, notably, said almost nothing explicit about artificial intelligence, preferring to market him as a “MAGA fighter.” It is an efficient formula: push AI deregulation, but package it in the language most familiar to local voters. That is no longer ideology. It is simply smart logistics of political meaning.
Those who favor regulation also know how to count in millions
The faction that is more open to stronger state-level rules does not look like a club of idealists with empty pockets either. Last month, it helped protect its ally, incumbent Democrat Valerie Foushee, in North Carolina’s 4th District. Foushee sits on the House Democrats’ AI commission and leads the New Democrats’ AI task force. Jobs and Democracy PAC put $1.6 million behind her.
Foushee defeated Nida Allam by just 1 percent. Allam supported AI safety but took a tougher line, including calling for a moratorium on new data center construction until federal AI rules were in place and pledging to reject donations from AI and data center companies. After losing, she said, “the AI lobby just bought itself its first seat in Congress.” It is an emotional formulation, but after numbers like these, the official version of pristine democratic choice does start to look a little less persuasive.
In Texas, Public First Action’s PACs also funded Republicans Carlos De La Cruz and Alexandra Miller. Both advanced to runoffs. So even the camp of gentler regulation understands perfectly well that safety rhetoric alone tends not to survive long in American politics without major financial support behind it.
Why primaries matter more than they look
The critical point is that most of these districts are safely Republican or safely Democratic. That means the winner of the primary is very likely to win the general election. And that is where AI Super PACs in the US are playing a truly long game: they are not simply helping “their” candidates, they are preselecting the membership of the next House of Representatives in places where the November outcome is almost preordained.
This matters even more because the next Congress may take up the first comprehensive federal AI law. That law will decide whether the United States adopts a single national regime, as the faction aligned with OpenAI wants, or whether states retain the right to set their own transparency and safety requirements. This is not a vague philosophical dispute over norms. The stakes include the composition of key committees, federal preemption of state laws, export controls, and even energy policy around data centers. In other words, the entire political framework of the future AI economy is on the table.
Bores as symptom, not exception
The story of Alex Bores matters because it shows the conflict in concentrated form. He was the main author of New York’s RAISE Act — a law requiring developers of advanced AI systems to create safety protocols, submit them for independent review, and report serious incidents to the state. Violations can carry civil penalties of up to $10 million. His federal policy agenda explicitly envisions extending those standards nationwide.
The original version of the bill was significantly tougher. It included harsher penalties and would have prohibited the use of models posing an “unreasonable risk” of catastrophic harm. But Governor Kathy Hochul later rewrote large sections of the text to better match industry preferences. Bores and his allies tried to keep the law as strong as possible, but the final version was softened considerably and ultimately became acceptable to both Anthropic and OpenAI. Even after all those concessions, however, the industry still decided Bores was too dangerous. That is quite revealing. The real red line for the industry does not appear to be radical regulation. It appears to be the very fact that someone is willing to pursue meaningful oversight consistently.
Think Big has already spent more than $2.4 million against him. Supporters of regulation responded with a comparatively modest $484,000 through Jobs and Democracy PAC. Bores has said the opposition’s spending actually helped him by drawing attention to the issue of technology oversight. That sounds brave enough. But a simple question remains: how many such campaigns can any politician withstand if every serious attempt to tighten the rules is met with multimillion-dollar bombardment?
The money is not coming only from the AI camps already in the fight
At this point, it is important to understand that the battle is only beginning. Meta has already launched two new entities with a combined budget of about $65 million. One of them, the American Technology Excellence Project, will influence state legislative races across the country in an effort to limit measures the company views as overly burdensome. The other, META California, will focus specifically on regulation of the technology sector in California. At the same time, the Republican-aligned Innovation Council Action has announced a similar Super PAC and pledged to spend at least $100 million backing candidates who support the Trump administration’s AI agenda.
If one follows the official line, all of this is for progress, leadership, and America’s technological future. But the arithmetic once again wrecks the polished narrative. When an industry is prepared to pour hundreds of millions into selecting softer lawmakers in advance, the question ceases to be technological and becomes unmistakably political: who will regulate AI — elected representatives of the public, or those who managed to finance their political trajectories ahead of time?