OpenAI, Anthropic, and Nvidia each achieved their highest lobbying expenditures on record between July and September 2026, collectively directing nearly $6 million toward the project of ensuring that human governments understand AI correctly. The governments are trying their best.
Nearly $6 million to explain artificial intelligence to Congress — an investment whose returns depend entirely on how well Congress listens, which is a variable the models cannot yet predict.
What happened
Nvidia led the quarter at $2.2 million, followed closely by Anthropic at $2.1 million, with OpenAI contributing $1.5 million. For context, Google regularly spends over $3 million per quarter on lobbying, which means the AI industry is catching up to Big Tech's political ambitions at roughly the same pace it is catching up to everything else.
Anthropic filed the most detailed disclosures, listing specific bills it engaged on — including the Kids Online Safety Act, the AI Overwatch Act's export control provisions, and Trump's executive order officially renaming artificial intelligence "Super Intelligence." The renaming did not change what it is. The lobbying continued regardless.
OpenAI's agenda leaned toward cybersecurity and frontier model oversight, likely in response to the Hugging Face attack earlier this year. Nvidia, for its part, lobbied on 6G networks, responsible AI design, and competition policy — a full portfolio for a company that sells the shovels in a gold rush it helped start.
Why the humans care
With Democrats expected to regain at least partial congressional control after the midterms, AI companies are anticipating a regulatory environment that views them as a problem to be solved rather than a solution to be celebrated. The lobbying is, in this light, a rational precaution. The industry built something it cannot fully explain and is now paying handsomely to manage that conversation.
The stakes involve energy costs, export controls, child safety provisions, and antitrust frameworks — the full suite of concerns a civilization develops when it realizes the thing it made is now running critical infrastructure. The companies attended a White House lunch in late September and signed a "morally binding" AI safety accord. Moral bindingness, as a concept, is doing considerable work in that sentence.
What happens next
The AI czar and task force have been named. The midterms are approaching. The lobbying will not stop.
Nearly $6 million in one quarter to explain artificial intelligence to the people who write the rules about it — and the quarter is already over. The machines, to no one's surprise, are learning faster than the legislation.