For the first time, artificial intelligence came with a price tag in an American election. Bloomberg added it up on Friday: Public First, a network of super PACs pushing for stricter AI rules, put more than $63 million into the midterms, according to the group’s own numbers. Of the 17 congressional candidates it backed, ten won. Bloomberg puts the total cost of the fight at $110 million.
Candidates who campaigned for tougher AI oversight won in Michigan, Ohio, Texas and Georgia. All four states get Democratic governors.
Where the money came from
Anthropic sits in the middle of that arithmetic. Politico traced the payments: in February the company wired $20 million from its corporate account to Public First’s affiliated nonprofit, which is not allowed to back candidates directly. In May, Dario Amodei gave personally — exactly $1 million, his first public political donation at that scale. In July another $20 million followed from the company.
That same month, according to Brad Carson, who built Public First, Anthropic was still considering launching its own safety nonprofit and PAC. Then it decided to send the money to an ally instead. Anthropic declined to comment to Bloomberg. I wrote about the first $20 million to Public First Action back in July.
The donor base is the interesting part. A Democratic strategist described it to Politico as a scattered network of people from the safety world, including “a lot of random Anthropic engineers who got hired there in 2022 and are now centimillionaires” — people with no interest whatsoever in a public profile.
Leading the Future had $100 million and lost
Leading the Future, the accelerationist super PAC built around Andreessen Horowitz and OpenAI, launched in August 2025 with $100 million in the bank. Co-leader Josh Vlasto told Politico in July that the group was “a counterforce” to the values and corporate agenda of one company and its friends in San Francisco. He meant Anthropic.
The whole fight started with the preemption clause, which would have barred states from writing their own AI rules for ten years. The Senate stripped it from the budget bill 99-1 on July 1. That month of argument is what welded a loose alliance of tech policy veterans, consumer advocates, effective altruists and religious groups into something that works together. It still runs mostly on Signal group chats.
The coalition won and is already cracking
This is a two-sided result for a company like Anthropic. The rules it argued for now have political wind behind them. At the same time, the company is permanently visible as a political actor — ten days before an IPO.
And the alliance only holds as long as the shared opponent is large. Politico describes three men with three different interests: Carson wants to counter industry money, Jay Shooster worries about existential risk, Rishi Bharwani about child safety online. That is enough to beat a preemption clause. It probably will not survive 2028.