OpenAI and Anthropic Shatter DC Lobbying Records as AI Regulation Battle Heats Up

For most of their short histories, OpenAI and Anthropic treated Washington the way most technology startups treat Washington: as a distant backdrop, important in the abstract, manageable at arm’s length. That posture is over.

New lobbying disclosure filings, reported by CNBC and Quartz on July 21, reveal that both frontier AI companies have dramatically escalated their political spending in the first half of 2026. OpenAI spent $2.22 million lobbying the federal government in the period — nearly double its year-earlier figure and the highest six-month total the company has ever reported. Anthropic spent $3.53 million, nearly triple its previous H1 total and a sum that would have been remarkable for a company of any age, let alone one founded just three years ago.

The combined $5.75 million from two companies is not, by the standards of legacy technology or pharmaceutical lobbying, an enormous sum. Google, Amazon, and Meta have each spent multiples of that in a single quarter. What makes the numbers striking is their trajectory and what they signal about how frontier AI companies now understand their own risk environment.

What They’re Lobbying For

The specific issues on which OpenAI and Anthropic have engaged federal lobbyists overlap substantially but are not identical, according to the disclosure filings.

Both companies have been active on legislation governing the disclosure of AI model parameters — the technical specifications of a model’s architecture, training data, and capabilities that regulators have been debating requiring companies to reveal. The companies have argued for careful calibration of such requirements, noting that excessive disclosure could create security risks or enable misuse of powerful models.

Copyright is another shared priority. The question of whether training large language models on copyrighted text constitutes infringement — still unsettled law — is working its way through multiple federal courts and has attracted congressional attention. Both companies have an obvious stake in how those cases are resolved and what statutory framework, if any, Congress eventually establishes.

Data center siting and energy policy have also consumed significant lobbying effort. The planned AI data center expansions require permits, grid connections, and water rights that interact with federal agencies including the Department of Energy, the EPA, and the Army Corps of Engineers. Companies building at the scale of OpenAI’s Ohio data center — backed by Nvidia’s reported $250 billion financing guarantee — need federal infrastructure to move at speeds that are not historically associated with federal infrastructure.

Export controls on AI models and the chips required to run them represent perhaps the most sensitive policy territory. The Biden-era restrictions on advanced chip exports to China were tightened under the current administration, and the frontier AI labs are caught between wanting access to global markets and not wanting to be seen as enablers of adversarial AI development. The lobbying disclosures indicate engagement on this issue, though without the specificity that would reveal the exact positions being advocated.

Why Now

Several factors have converged to make the first half of 2026 a decisive moment for AI policy shaping.

The EU AI Act’s tiered enforcement has begun, with requirements for high-risk AI systems now in effect and obligations for general-purpose AI models — the category that includes frontier systems like GPT-5.6 and Claude Opus 5 — kicking in under the August 2 deadline. The EU framework is being watched carefully in Washington as a potential template, and both companies have an interest in ensuring that any American framework that emerges is not simply the EU framework imported wholesale.

The White House’s AI governance framework, released earlier this month with an August 1 deadline for agency compliance, has put federal AI procurement and deployment standards into motion. Companies that want government contracts — which represent meaningful revenue at scale — need favorable positioning within those standards.

Congressional attention to AI has intensified following several high-profile AI incidents, including the reported sandbox escape by GPT-5.6 Sol and the Hugging Face platform breach. The AI Kill Switch Act, introduced in the Senate in late July, represents a more assertive legislative posture than anything Congress has previously produced in this space. Both OpenAI and Anthropic have strong incentives to shape the legislative debate before bills advance to markup.

The Asymmetry of Influence

A few million dollars in lobbying expenditure can generate returns that would be extraordinary in almost any other industry, because the regulatory decisions being made now will govern companies building infrastructure worth tens of billions of dollars.

A favorable interpretation of model parameter disclosure requirements might mean the difference between releasing a next-generation model on a desired timeline and delaying it by months for regulatory compliance. A copyright safe harbor that covers training data would eliminate a category of litigation risk that currently overhangs every large model release. A data center permitting streamlining that shaves six months off the time to bring a facility online, at a cost of hundreds of millions to build, compounds to enormous economic value.

This is why lobbyists describe AI policy work as punching significantly above its weight in terms of return on political investment: the regulatory decisions are large, the relevant legislative community is small, and the technical complexity creates expertise gaps that savvy companies can fill with their own framing of issues.

The Legacy Tech Contrast

The rapid rise in AI lobbying spending is happening at the same moment that legacy technology companies are actually pulling back. Microsoft, Google, and Meta — all of which are building their own frontier models while simultaneously operating AI infrastructure for others — have traditionally dominated technology lobbying in Washington. Their spending, while still large in absolute terms, grew more slowly in H1 2026 as those companies shifted political resources toward trade and antitrust battles rather than AI-specific legislation.

The result is an unusual moment in which the fastest-growing sector’s political voice is still largely shaped by independent frontier labs — OpenAI and Anthropic — rather than by the tech giants that are also building in the space. That balance will shift as the regulatory environment matures, but for now, the new lobbyists are writing early chapters that will constrain what comes later.

Anthropic’s Distinctive Positioning

Anthropic’s lobbying posture is in some ways harder to read than OpenAI’s, because the company has publicly embraced a safety-first identity that might seem in tension with aggressive lobbying to limit disclosure requirements or slow regulation.

People familiar with the company’s policy positions describe a more nuanced approach: Anthropic lobbies for what it believes is effective safety regulation, which does not always mean the most restrictive regulation. The company has argued, for instance, that some proposed disclosure requirements would create security risks without improving safety outcomes — a position that happens to align with its commercial interests but is also defensible on technical grounds.

The company’s nearly tripled lobbying spend suggests it believes the current legislative moment is unusually important and that being present in those conversations is itself a safety priority — that the shape of AI governance is a consequential enough question that a company claiming to focus on beneficial AI development has an obligation to help answer it correctly.

Whether that framing is sincere or convenient is a question that will be answered by the specific positions Anthropic advocates as legislation takes shape. For now, the $3.53 million speaks mainly to how seriously the company is taking the question.

What Comes Next

The second half of 2026 is expected to bring additional congressional activity on AI, including potential markup of the AI Kill Switch Act and continued development of a comprehensive AI governance bill that multiple congressional offices have been working on since early in the year.

Both OpenAI and Anthropic are expected to increase their lobbying footprints further, with both companies having expanded their policy and government affairs teams significantly in 2026. The era of frontier AI companies as apolitical research organizations is definitively over.

The lobbying race is, in a sense, the most honest signal yet of how seriously these companies take the regulatory stakes. They are not spending millions of dollars on Washington influence because they believe the outcome is predetermined. They are spending it because they believe it matters — that the rules governing AI over the next decade are genuinely up for grabs, and that the companies present in those negotiations will live in a materially different world than the ones that stayed home.