The quiet war for AI's regulatory future just got a $40 million megaphone.
The Summary
- Anthropic doubled down with another $20 million to Public First Action, bringing total commitment to $40 million ahead of November's midterms
- CEO Dario Amodei personally kicked in $1 million to an allied Super PAC, signaling this isn't just corporate strategy
- Public First Action is a 501(c)(4) nonprofit that doesn't disclose donors but can funnel money to ads highlighting specific candidates on AI policy
- This sets up a proxy battle with OpenAI president Greg Brockman's competing super PAC over who shapes AI regulation
The Signal
Anthropic just made the biggest corporate bet yet that AI regulation will be written by whoever shows up with the most money in 2026. The $40 million commitment dwarfs typical tech lobbying spend on single-issue campaigns. For context, the entire crypto industry spent roughly $130 million across all races in the 2022 midterms. Anthropic is approaching one-third of that number, alone, for a single policy vertical.
Public First Action is structured as a dark money nonprofit, meaning it can accept unlimited donations without naming sources. But it can route funds to allied Super PACs that do run candidate-specific ads. The first ad highlighted Tennessee Republican Marsha Blackburn's AI safety work. The pattern is clear: find lawmakers who talk about AI risk, amplify them, make AI regulation a campaign issue before November.
"We need policymakers and candidates to put forward measures that mitigate risks."
The timing matters. Anthropic specifically cited Claude Mythos, their latest model release, as raising urgency. Translation: as capabilities accelerate, the window to shape rules before something breaks is closing. They're not waiting for a precipitating event. They're buying influence while the policy landscape is still forming.
Key strategic moves:
- Structured as nonprofit to avoid donor disclosure requirements
- Built bipartisan leadership with former GOP Rep Chris Stewart and Democrat Brad Carson
- Targeting down-ballot races, not just federal headliners
- Personal CEO donation signals this transcends corporate comms strategy
This isn't lobbying in the traditional sense. Traditional lobbying is quiet, transactional, focused on specific bills. This is building a political constituency for AI safety regulation from scratch. The bet: if enough candidates campaign on AI risk mitigation, whoever wins will feel obligated to deliver.
The elephant in the room is who Anthropic is fighting. The article mentions a competing super PAC backed by OpenAI's Greg Brockman, but doesn't detail the clash. Safe bet: the fight is over prescriptive safety requirements versus industry self-regulation. Anthropic has consistently pushed for compute thresholds, third-party auditing, and reporting requirements. OpenAI has pushed for voluntary commitments and frontier model forums. Both companies want regulation. They disagree violently on what kind.
The Implication
Watch which candidates Public First Action backs in the next three months. That's your roadmap for what AI safety regulation will look like in 2027-2028. If they're funding challengers, not just incumbents, this is about changing Congress, not just influencing it.
For anyone building in the agent economy: regulatory capture is happening in real time. The companies defining AI safety today will define what you're allowed to deploy tomorrow. If you're not tracking which specific bills Blackburn and Public First-backed candidates are drafting, you're flying blind. The compliance burden is being written right now, and it won't be optional.