OpenAI and Anthropic broke their own lobbying records in Q2 2026, together spending $3.17 million on federal influence between April and June, a 23% jump from the prior quarter. Smaller AI companies have no seat at that table and will inherit whatever rules these two help negotiate.
Anthropic spent $1.97 million on Washington lobbying in the second quarter of 2026. That's its biggest quarter ever, up 26% from Q1, and it already exceeds the $3.1 million the company spent in all of 2025. OpenAI was not far behind at $1.2 million, up nearly 18% from the prior quarter. As CNBC reported, both companies more than doubled their year-over-year spend from Q2 2025. That kind of acceleration does not happen without a reason: the regulatory window is open, and they know it.
The issues on their lobbying disclosures read like a checklist of everything that will determine how the AI industry is built: cybersecurity, copyright, cloud procurement, export controls, AI safety standards. In the second quarter, Anthropic lobbyists met with officials at the White House, the Commerce Department, and the Treasury - plus members of both chambers. That's a lot of rooms. That's not a company filing paperwork. That's a company trying to shape outcomes.
This week offered a vivid illustration of what those meetings are for. On July 23, Representatives Ted Lieu and Nathaniel Moran introduced the AI Kill Switch Act, a bipartisan bill that would require AI companies to maintain the ability to shut down, throttle, or suspend their models, and give the Department of Homeland Security new power to order those shutdowns. The bill was triggered in part by a frontier model security incident involving OpenAI. If you are a company whose core product could be switched off by a federal agency, you want to be the person who helped write the definitions of when that switch gets flipped. Both OpenAI and Anthropic do. Neither has a lobbyist who is going to miss that conversation.
Zoom out and the broader picture is striking. According to Issue One, 11 major tech and AI firms collectively spent roughly $41 million lobbying Washington in the first half of 2026, about $226,000 per day. Meta led the pack at nearly $6 million in Q2 alone; Alphabet spent $5.3 million. Anthropic, remarkably, outspent Nvidia in the second quarter. A semiconductor company with a $3 trillion market cap got outlobbied by a five-year-old AI lab. That tells you how fast these priorities are shifting.
What founders who can't afford K Street are left with #
Here's the thing about lobbying arms races: the losers don't usually know they've lost until the rules are already written. The concerns are not hypothetical. At the Axios AI+NY Summit in June, startup founders said plainly that they feared new AI regulations would entrench incumbent advantages and foreclose competition before smaller players could get established. Compliance costs that cost a rounding error for a company valued at $157 billion, OpenAI's last reported valuation, can be existential for a 20-person team. Copyright rules that favor model builders who can afford litigation. Export control frameworks that create licensing burdens. These are the kinds of outcomes that emerge when the companies being regulated are the same companies staffing the meetings that produce them.
The asymmetry runs deeper than money. OpenAI and Anthropic are not just writing checks; they're deploying sophisticated government affairs operations built on relationships that most startups don't have and can't build in the time it takes Congress to move a bill. Anthropic alone has spent more than $3.5 million in the first six months of this year. A seed-stage company's entire annual burn rate, gone on a single quarter of lobbying. The founders building the next generation of AI tools will not get to vote on the regime they inherit, but they will absolutely have to live inside it.
Frankly, none of this is surprising. Every major technology wave has produced a moment where the companies who got there first tried to pull the ladder up behind them, not always deliberately, but effectively. The result in this case will be an AI regulatory framework shaped overwhelmingly by two companies with specific interests in specific outcomes, while the thousands of developers and startups building on top of these models watch from outside. If Congress passes something like the AI Kill Switch Act, or moves on copyright liability for training data, or codifies export control rules on Chinese models, the fine print will largely reflect whoever showed up to the meetings. Right now, that's Meta, Alphabet, OpenAI, and Anthropic. The rest of the industry can read the final bill when it comes out.
Also read: Twenty-one banks just lined up behind SoftBank's $40 billion OpenAI bet • Etched raises $300M at $10.3B valuation as Sequoia backs the Harvard dropouts it once rejected • CXMT surges 500% on Shanghai debut but its $490 billion valuation outpaces what the company can actually build