Anthropic and OpenAI Just Set Lobbying Spending Records

Federal lobbying disclosures for the second quarter of 2026 came out this week, and they show something worth pausing on: Anthropic and OpenAI just spent more money lobbying the US government than either company ever has before. Anthropic reported $1.97 million for the April-through-June quarter, up 26% from Q1. OpenAI reported $1.2 million, up 18%. Combined, that’s $3.17 million in three months, roughly double what the two companies spent over the same stretch in 2025.

A couple of comparisons make the scale of this land. Anthropic’s quarterly spend now exceeds Nvidia’s ($1.25 million) and nearly matches Oracle’s ($2 million) — two companies with vastly larger revenue and a much longer history of managing Washington relationships. OpenAI came within $50,000 of Nvidia too. Add Anthropic’s first-quarter filing and its 2026 half-year total already tops $3.5 million, more than the $3.1 million it spent for the entirety of 2025. These are AI labs that, a few years ago, barely registered on lobbying trackers next to legacy tech and defense contractors. Now they’re outspending some of them.

What are they lobbying about? The disclosures list cybersecurity, copyright, cloud computing, and defense procurement as focus areas. That’s a broad spread, but it maps onto real pressure points. Copyright lawsuits over training data keep piling up. Cloud and defense procurement rules determine who gets to sell AI services to the federal government, and at what margin. Cybersecurity policy — including export controls on frontier models — has become an existential concern for labs whose products can be restricted overnight by a single agency directive.

That last point isn’t abstract for Anthropic. In mid-June, the Commerce Department ordered the company to cut off foreign access to its newly launched Fable 5 and Mythos 5 models, citing national security concerns after a jailbreak was found that could turn the models into unrestricted cyber tools. The shutdown lasted about two weeks before Commerce allowed a limited restoration. Whatever the merits of that specific order, it’s a vivid demonstration of how much leverage a single regulatory decision has over a frontier AI company’s business — and a very concrete reason to want more input into how those decisions get made. It’s not a stretch to read the 26% jump in Anthropic’s Q2 spending, filed for the same quarter the shutdown happened, as at least partly a response to that experience.

Why this deserves more attention than a line item in a filing. Lobbying spending is a leading indicator of how much a company thinks government decisions will affect its bottom line, and by that measure, AI labs now see Washington as roughly as consequential to their business as Nvidia or Oracle do — companies for whom procurement contracts and export rules have long been central. That’s a real shift from a year or two ago, when AI companies mostly framed themselves as technology developers rather than regulated entities managing relationships with regulators. We’ve written before about OpenAI’s proposal to give the federal government an equity stake in the company — a more dramatic version of the same underlying dynamic, where a lab wants to convert its regulator into something closer to a stakeholder. Record lobbying spending is the more conventional, and much more common, version of that same move: pay to have a seat at the table where the rules that can shut down your product, or hand you a contract, actually get written.

It’s also worth being honest about what these numbers don’t tell you. Federal lobbying disclosures capture registered lobbying activity — meetings, calls, and communications by registered lobbyists — not the full universe of ways a company can shape policy, which includes think-tank funding, campaign contributions through PACs, executive testimony, and informal access that doesn’t require registration at all. A $3.17 million combined quarterly total sounds large next to what AI labs spent two years ago, but it’s still modest next to what, say, the pharmaceutical or oil and gas industries spend lobbying on regulations that affect them just as directly. The interesting part isn’t that the absolute number is enormous — it’s the trajectory. Two industries that used to treat Washington as an afterthought are now budgeting for it like industries that have been shaped by regulation for decades, because they’ve each had a direct, expensive lesson in just how fast that shaping can happen.

None of this means AI regulation is now settled or captured — Congress still hasn’t passed comprehensive federal AI legislation, and most of the actual authority being exercised over these companies so far has come from existing statutes (export control law, in Anthropic’s case) rather than anything built for AI specifically. But the spending pattern tells you where these companies expect the next fights to happen, and how seriously they’re taking the possibility of losing them. Worth watching in the next quarter’s filings: whether this was a one-time reaction to a rough few months, or the new baseline.