AI Lobbying Record Q2

Anthropic spent $1.97 million on federal lobbying last quarter. That's more than Nvidia.

Let that sit for a second. A company that didn't exist five years ago now outspends the most valuable chipmaker on Earth in Washington.

The Q2 filings landed yesterday. Anthropic: $1.97M, up 26% from Q1. OpenAI: $1.2M, up 18%. Combined, $3.17M for the quarter — a 23% jump. Anthropic has already spent more in the first half of 2026 ($3.5M) than it did in all of 2025 ($3.1M).

The line items tell you more than the totals. Export controls. Cybersecurity standards. Copyright. Cloud computing. Defense contracts.

Here's what I think most people are missing: this isn't a story about influence. It's a story about cost structure.

When your model vendor is spending millions to shape the rules it operates under, regulation has stopped being a background condition and become an active input to their roadmap. Which means it's an active input to yours. The capability you're building on today exists partly because of a policy fight being waged right now — and it can be constrained by one just as fast.

We saw the preview of this in June, when a government action pulled flagship models offline and enterprise workloads went dark overnight. That wasn't a technical failure. It was a policy failure that landed on customers.

The practical takeaway for anyone running AI in production: your vendor risk assessment needs a regulatory column. Not "are they compliant" — that's table stakes. Ask which policy outcomes their business depends on, and what happens to your workload if those go the other way.

Model quality is converging. Policy exposure isn't.

What does your vendor evaluation look like today — and does regulatory dependency show up anywhere in it?