The price of writing an AI bill is now $7.6 million

Roughly $21 million of AI-industry money has landed on a single Manhattan House primary. The contest is between two factions of capital, one aligned with OpenAI, one with Anthropic, and the load-bearing question is which lab's preferred regulatory worldview gets a friendlier Congress. The seat matters. What matters more is what the spending tells every state legislator in the country about the cost of writing an AI bill.

The facts, briefly. Alex Bores is a New York State Assemblymember running in the June 23 Democratic primary for NY-12, the seat Jerry Nadler is vacating. A super PAC called Leading the Future, funded by OpenAI president Greg Brockman, Marc Andreessen, and Joe Lonsdale, has spent about $7.6 million attacking him. Groups affiliated with Anthropic and its investors have spent over $10 million supporting him, and Ripple co-founder Chris Larsen has pledged a further $3.5 million.¹² Bores's offence, in the eyes of Leading the Future, is that he wrote and passed AI safety legislation in Albany.¹

That is the surface. Below it is a more useful story.

Who is actually in the room. Two groups of AI-industry funders are arguing, at considerable expense, about the shape of federal AI policy. The workers whose jobs are being restructured by the technology these funders are building are not contributing to either PAC. Neither are the patients whose clinical encounters are being mediated by AI scribes, the students whose essays are being graded by language models, the renters whose applications are being filtered, or the welfare claimants whose benefits decisions are being routed through automated systems. The contest is internal to the industry. Everyone else is watching.

This is the first distributional fact about NY-12, and it should be named before anything else. The regulatory framework that emerges from this Congress will set the terms under which AI is deployed into hospitals, schools, workplaces, and benefit offices for years. The people who will live inside that framework have no $7.6 million PAC. The people building the systems have two.

The chilling effect is the product. Bores has said the spending is designed to deter other state legislators from touching AI, and on this he is almost certainly right. $7.6 million against one assemblyman is not a rational expenditure if the only goal is winning NY-12. The seat could be contested for a fraction of that. What the spending buys is not the seat. It buys a number — a publicly legible figure that every state representative in Albany, Sacramento, Lansing, and Austin can now look up.

The number says: if you introduce AI legislation in your state house, expect this. Expect a super PAC with an anodyne name to materialise in your next primary. Expect $7 million of attack ads on local television. Expect mailers to every household in your district questioning your judgement on issues unrelated to AI. Expect to spend the next eighteen months of your life raising defensive money instead of legislating.

For a state legislator earning a part-time salary, contemplating a bill that might attract this attention, the calculation is straightforward. Most will decide their constituents' interests are better served by working on something else. That is the demonstration effect, and it is the actual deliverable. Bores winning or losing is almost incidental to it.

"Pro-regulation" as a piece of language. Leading the Future publicly describes itself as supporting AI regulation.² It is spending $7.6 million against the only candidate in the race who has actually written and passed any. This is worth dwelling on, because it is the cleanest example I have seen this year of how regulatory framing is being weaponised in AI policy.

The move is not new. Every regulated industry eventually learns to say it supports the right kind of regulation while opposing every specific instance of it. What is new is the scale and the speed. AI is six years into mass deployment and the industry has already built the rhetorical apparatus that took tobacco forty years and finance most of a century. The PAC is funded. The framing is set. The candidate who actually legislated is the target.

The PAC opposing the only candidate who has written AI law describes itself as pro-regulation. The framing is the regulation.

The steelman, taken seriously. There is a real argument for Leading the Future's position, and it deserves engagement rather than dismissal. State-by-state AI regulation does create compliance fragmentation. Fifty different definitions of automated decision-making, fifty different audit regimes, fifty different liability standards — this is a genuine problem, and it falls hardest on smaller developers who cannot afford fifty compliance teams. A single federal standard would be cleaner. Bores's law, replicated in forty-nine variants, could produce a regulatory environment that is bad for safety and bad for deployment at once.

I take that seriously. It is the argument I would make if I were running Leading the Future's communications. The trouble is that the people making it have not spent $7.6 million advocating for a federal standard. They have spent $7.6 million attacking a state legislator. If the concern were genuinely about fragmentation, the spend would be on lobbying Congress to pass a preemptive federal bill that actually constrained the industry. The spend is instead on ensuring that the legislator who tried to constrain the industry at the state level does not get to constrain it at the federal level either. The pattern reveals the preference.

Anthropic is not the answer to OpenAI. The temptation, given how the race is being covered, is to read this as a fight between a responsible AI company and an irresponsible one, with Bores as the responsible side's candidate. That reading is too clean and ORA should not endorse it.

Anthropic has its own commercial interest in the outcome here. The company has publicly supported federal AI rules that would likely preempt state legislation, which is convenient for any large incumbent. Federal preemption favours the labs that can already afford a federal compliance operation; it disadvantages newcomers and smaller players who could have lived with a lighter state regime. Anthropic supporting Bores is not the same as Anthropic supporting the public interest. It is Anthropic supporting a regulatory architecture that happens to suit its position in the market.

The honest reading is that two large AI companies, with different competitive positions, are funding opposite sides of a primary to advance their respective regulatory preferences. Both preferences are defensible. Neither is identical to what would serve the people most affected by AI deployment. The question of what those people would want, if asked, is not on the ballot, because they are not the ones funding it.

Bores himself. I have not read his New York bill closely enough to tell you whether it is good legislation. The reporting frames it as one of the strongest state AI safety laws in the country, which tells me about its political profile more than its distributional content. A law can be strong against industry and still leave the workers, claimants, and tenants on the receiving end of AI systems with thin protections. Supporting Bores because OpenAI's funders oppose him is the laziest possible form of political reasoning, and I am not going to do it.

What is defensible to say is narrower. A state legislator wrote a bill, passed it through a real legislature, and is now facing $7.6 million in attack spending from the industry the bill touched. Whether his specific bill is the right model for federal AI policy is a separate question from whether legislators should be able to write AI bills at all without ending their careers. The second question is the one the spending answers, and it answers it badly.

What the $21 million actually tells us. AI companies do not spend $21 million on a House primary unless they believe Congress will materially shape their business. The size of the bet is itself information. The industry's internal estimate of regulatory risk is high — higher, perhaps, than the public discourse around AI inevitability would suggest. Companies do not fund proxy wars over outcomes they consider settled. They fund them over outcomes they think can still go either way.

That is the one piece of good news in this story. The fight over AI governance is not over. The industry would not be spending this kind of money if it were. The bad news is that the fight is being conducted between factions of capital, in a primary most Americans will not vote in, over a seat in a district most Americans cannot find on a map. The terms on which AI gets deployed into your workplace, your clinic, and your child's classroom are being negotiated right now, and the negotiation does not include you.

Bores may win on Tuesday. He may lose. Either way, the price of writing an AI bill has been set, in public, by the industry being regulated. Every state legislator in the country now knows what it is.

Footnotes

¶