OpenAI and Anthropic Hit $3.17M Q2 Lobbying Record as AI Policy Moves to Washington
OpenAI and Anthropic set new federal lobbying records in Q2 2026, spending a combined $3.17 million from April through June. That is up 23% from Q1 and roughly double their year-earlier pace.
The numbers are still small beside the largest regulated industries. The direction is not. Frontier AI labs are becoming Washington operators because model releases, data centers, copyright, export controls, defense procurement, and cybersecurity are now policy questions, not just product questions.
The Spend
| Company | Q2 2026 lobbying | Quarter change |
|---|---|---|
| Anthropic | $1.97M | +26% |
| OpenAI | $1.20M | +18% |
| Combined | $3.17M | +23% |
Anthropic’s first-half 2026 lobbying total is now above $3.5 million, already higher than its full-year 2025 spend of about $3.1 million. OpenAI is moving in the same direction, with Q2 spending just under Nvidia’s $1.25 million for the quarter.
The broader tech lobbying machine remains much larger. Meta, Amazon, Google, Microsoft, and Nvidia spent a combined $21.25 million in Q2, roughly flat with Q1. Meta led that group at $5.99 million, followed by Amazon at $4.36 million.
What Changed
AI lobbying used to be framed around abstract safety and innovation language. The current docket is more concrete. Filings now point to cybersecurity, copyright, cloud computing, defense procurement, export controls, model oversight, and data center infrastructure.
That list maps directly to the pressure points of the industry. Labs need power approvals and compute access. They need federal preemption or harmonisation to avoid a patchwork of state AI rules. They need copyright exposure contained. They need defense and intelligence contracts without being boxed into procurement or supply-chain restrictions. They need export controls that slow China without cutting off revenue or accelerating Huawei.
The Anthropic Split
Anthropic’s spending pattern is especially useful because it does not line up cleanly with the usual “industry against regulation” story. The company has supported stricter model oversight in several debates while also fighting for access, procurement eligibility, and influence over the exact shape of those rules.
That is the new posture of frontier labs: regulate, but regulate in a way that preserves their own operating model. OpenAI, Anthropic, Google, Meta, Microsoft, and Nvidia can disagree loudly on open weights, safety thresholds, and state preemption while still sharing the same underlying objective, keeping Washington close enough to shape outcomes before they harden into law.
Q2’s $3.17 million is not a ceiling. It is the point where frontier model policy became a standing cost of doing business.