AI companies should stop pretending midterm spending is neutral

OpenAI- and Anthropic-linked spending in the midterms is lobbying, not neutrality.

AI companies should stop pretending their midterm spending is neutral public-interest advocacy. When Anthropic puts $20 million into a related nonprofit that openly fights federal efforts to freeze state policy, that is a political bet with a corporate objective, not a civic gesture. The same logic applies to any OpenAI-aligned spending aimed at shaping the rules that govern model deployment, liability, and state-by-state oversight.

These donations are about power over the rulebook

Regulation is not an abstract policy debate for frontier AI labs. It determines whether states can move faster than Congress, whether safety disclosures become mandatory, and whether companies face real consequences when their systems cause harm. A contribution of $20 million is not symbolic in that context. It is a bid to influence who writes the rules and how much friction those rules create.

The clearest evidence is in the stated purpose of Public First Action: it “opposes federal efforts to freeze state progress without adequate federal safeguards.” That phrasing is carefully chosen, but the intent is obvious. The money is meant to preserve a policy environment where AI firms can keep operating under a patchwork rather than a stricter national regime. That is a strategic preference, and it should be labeled as such.

Opacity makes the spending more corrosive, not less

AI labs already enjoy an unusual amount of public trust for companies with enormous technical leverage and limited external oversight. When political spending flows through related nonprofits, that trust erodes further because voters cannot easily see who is paying, who is coordinating, and what exact policy outcomes are being purchased. If the public has to reverse-engineer the influence campaign from nonprofit filings and press reports, the disclosure regime is failing.

There is a simple comparison here: if a company wants to argue for a policy, it should do so directly and transparently. A direct statement is accountable. A network of related entities is harder to trace and easier to sanitize. In a field as consequential as AI, that distinction matters. The more powerful the technology, the less patience the public should have for political spending that hides behind soft language about safeguards and progress.

The counter-argument

Supporters of this spending will say the labs are defending innovation against blunt federal overreach. That argument is not frivolous. AI systems are already subject to emerging state rules, and a chaotic patchwork can punish smaller developers while rewarding the largest incumbents who can afford compliance armies. From that angle, political spending is a defensive move to keep the market from fragmenting before federal standards exist.

They will also argue that these companies are among the few actors with enough technical knowledge to warn lawmakers about bad regulation. If legislators write rules without understanding model behavior, training costs, or deployment risks, the result can be performative policy that neither protects the public nor helps builders. In that sense, spending is framed as expertise entering the process, not corporate capture.

That defense only goes so far. Expertise does not erase self-interest, and self-interest does not become public service just because the issue is complicated. The right response is not to ban companies from participating in politics. The right response is to require clean disclosure, direct advocacy, and a much higher standard of transparency when firms with foundational AI products spend to shape the law. If these labs believe their position is sound, they should be willing to defend it in the open.

What to do with this

If you are an engineer, PM, or founder, treat political spending as part of your company’s product strategy, because regulators and users will. Ask who is funding the advocacy, what policy outcome it seeks, and whether your public stance matches your internal risk posture. If it does not, fix the mismatch before it becomes a trust problem. In AI, governance is now part of the stack, and pretending otherwise only makes the eventual backlash worse.