Two labs, one venue: the AI lobbying money tells buyers to plan for state rules

Both playbooks route through the same buildings. OpenAI's "reverse federalism" asks state legislatures to pass consistent AI laws so that a national standard emerges from the bottom up, on the argument that developers should not have to clear 50 separate compliance regimes ^[6]. Anthropic is working the same legislatures for stringent safety mandates ^[8]. The two labs disagree about how hard the rules should bite, not about where they get written. A compliance plan built on the assumption that Washington will eventually preempt the states is a plan neither of the best-funded participants is currently buying.

The disclosed federal figure is also the wrong figure to anchor on. Anthropic's $1.97m is 62 percent of the $3.17m combined total ^[12], leaving OpenAI at roughly $1.2m for the quarter ^[11], so Anthropic outspent its rival by about two thirds ^[15]. Against that, the $20m Anthropic sent to Public First Action in July ^[9] is about 6.3 times the entire combined quarterly lobbying number that gets quoted ^[13], and roughly ten times Anthropic's own share of it ^[14]. The lobbying line item is the visible part. The advocacy vehicle is where the weight sits.

That matters for anyone writing a two-year vendor policy, because the safety-mandate side of the argument is the side with the larger cheque and the earlier start: Anthropic opened its Washington office in April 2026, a month before OpenAI ^[3][4], and in 2025 became the first major lab to endorse California's AI regulation law ^[7]. The report does not say what obligations that California law imposes, so treat the specific line items as unknown. The direction is not unknown. If stringency wins in two or three large states, the strictest applicable state standard becomes the practical floor for everyone selling into those markets, and the uniformity OpenAI is asking for arrives as convergence on that floor rather than relief from it.

The incentives behind each position are not subtle, and the source states them plainly: OpenAI has the largest user base and the most deployed products, so a patchwork of strict state laws is more expensive for it than a single national rule ^[19], while Anthropic's safety-research positioning means stricter mandates raise the bar for competitors on a test it can argue it already passes ^[20]. Both companies are reportedly preparing for possible IPOs, and regulatory uncertainty unsettles institutional investors ^[16]. Pre-IPO issuers want that uncertainty resolved before a book is built, not after, which argues for rules landing sooner. Midterm elections, which can redirect tech legislation entirely ^[17], set the clock on that.

There is a comforting read available. The combined $3.17m is modest next to what the largest tech companies deploy ^[18], and the 23 percent quarterly increase ^[1] works out to roughly $590,000 in absolute terms over a prior-quarter base near $2.58m ^[10]. Small money, small stakes. The less comforting read is that Anthropic's quarterly total already exceeded Nvidia's for the same period ^[2], and that the real spending has moved off the lobbying disclosure and into a political group ^[9]. Buyers should assume the second read, and assume the rules being drafted this way will be state rules.