AI companies spend record sums on Washington lobbying

America’s leading AI companies spent record sums on lobbying in Washington in recent months as they try to limit unfriendly regulation and shape the US rules around data centres and vetting advanced models.
OpenAI nearly doubled its federal lobbying expenditure to a record $2.22mn in the first half of 2026, compared to last year, while Anthropic nearly tripled its spending to $3.53mn, according to federal disclosures. Google, along with Microsoft, spent as much as in any quarter since 2020.
The increase comes as the AI industry confronts political battles on multiple fronts and builds a more permanent influence operation in Washington, according to lobbyists and policy experts.
“The lobbying offensive has been as much about deterring regulation as making the case for an affirmative government industrial policy that supports the industry,” said Amba Kak, co-executive director of AI Now Institute and senior adviser on AI to the Federal Trade Commission during Joe Biden’s administration.
The top issues on the lobbying agenda include the federal rules governing the release of new models, after the US last month wielded export controls to stop the rollout of Anthropic’s Fable model over cyber security concerns.
Companies also want to influence policy around the construction and power for massive data centres, as well as rules governing whether developers make the underlying parameters for their models publicly available, according to federal disclosures and lobbyists involved in the policy discussions.
Joseph Hoefer, chief AI officer at Monument Advocacy, said: “It’s an export-controlled technology like semiconductors, an infrastructure build-out like telecoms or power, a procurement relationship like defence, and a liability question like the [social media] platforms, all at the same time. No single playbook covers that.”
The industry “lost loudly” an effort to get Congress to ban state-level AI regulation, something Donald Trump has tried to enforce with an executive…