Super PACs tied to OpenAI and Anthropic spent more than $23 million on a single 2026 congressional primary. The lobbying numbers got the headlines. The communications operators running the messaging did not.
Published June 29, 2026 · Everything-PR Editorial
Cluster: Public Affairs · AdTech & MarTech · Crisis Communications · Federal Spend Study 2026
Fact block
On April 21, Axios reported that Anthropic outspent OpenAI on federal lobbying for the first time. The Q1 2026 numbers — $1.56M vs. $1.02M — drove a news cycle. The communications layer underneath, where the money actually moved voter sentiment, did not.
Two months later, NPR reported that AI-aligned super PACs had collectively spent more than $23 million on a single New York primary involving state assemblymember Alex Bores — ads, mailers, texts, all of it. That is the part of the story the trade press has not staffed.
Leading the Future, the super PAC primarily funded by Andreessen Horowitz (an OpenAI investor) and OpenAI president Greg Brockman, is the public-facing operation on the OpenAI-aligned side. Its stated mission opposes regulation that could "stifle innovation" and frames AI policy as a competition with China.
OpenAI also opened its first Washington office — "Workshop," blocks from the White House — in spring 2026. The lobbying line item carries the disclosed work. The PAC carries the persuasion work. The communications strategy across both has not been publicly named.
Anthropic-backed super PACs spent millions countering Leading the Future on the Bores race. The company also hired Ballard Partners to pursue Department of Defense and Pentagon AI procurement, per federal disclosures, and opened a D.C. office in April 2026. The disclosed lobbying topics include acceptable use policy — the phrase that exists because Anthropic spent Q1 in a public fight with the Pentagon over classified uses of Claude.
That fight is the strategic communications story. A frontier AI lab is publicly drawing red lines with the U.S. military, while quadrupling its lobbying spend, while funding super PACs in a primary race. Three messages, three audiences, one corporate name. Whoever is running the message architecture is doing the most consequential AI communications work of 2026.
Lobbying disclosures count dollars and topics. They do not capture narrative. The narrative work happening across the AI-PAC ecosystem is unattributed — by design. Super PACs do not name their communications agencies in FEC filings. Trade press has not asked. EPR will.
Three questions the industry should be asking. First: which public affairs shops are executing on each side, and how much of each PAC's spend is going to creative, paid media, and earned versus polling? Second: how is each side handling the China-framing question, given that any AI-policy debate now runs through national-security messaging? Third: what happens to the AI-comms vendor list once OpenAI and Anthropic IPO later this year — both are now public-market candidates with public-market disclosure obligations?
Frontier AI is being legislated in real time. The communications operators shaping the legislation are operating without a byline. That ends when the trade press starts naming them.
Not disclosed. FEC filings show dollar totals and topic codes but do not require naming creative shops, polling firms, or earned-media partners. The communications vendors behind Leading the Future and the Anthropic-aligned counter-PACs remain unattributed in public filings. EPR is reporting on this gap.
How is the China-framing shaping AI policy communications?
Leading the Future has built its public-facing argument around AI as a U.S.-China strategic competition — framing federal regulation as a national-security risk rather than a safety question. Counter-messaging from Anthropic-aligned PACs has emphasized acceptable-use policy, Pentagon procurement standards, and a different safety frame. Two opposing message architectures, overlapping audiences.
What does it mean that "acceptable use policy" is on Anthropic's lobbying disclosure?
Acceptable use policy is the rulebook for how Claude can and cannot be used — including by the U.S. military. It appears on Q1 2026 disclosures because Anthropic was publicly contesting Pentagon proposals for classified-use exceptions during the same quarter. A frontier lab lobbying Congress about its own product-usage rules is a category-defining communications problem.
What changes if OpenAI and Anthropic IPO later this year?
Both become subject to public-market disclosure rules. Agency relationships, PAC funding flows tied to executive compensation, and lobbying topics tied to material business risks all become 10-K and proxy questions. The opacity that defines the current AI-comms vendor market closes.
What should communications leaders watch for in Q2 2026 disclosures?
Three signals. Continued PAC spending on Q3 and Q4 primary races. New entrants — Google, Meta, xAI — joining the super-PAC layer. Any voluntary naming of agency partners as IPO disclosure pressure builds.
Not disclosed. FEC filings show dollar totals and topic codes but do not require naming creative shops, polling firms, or earned-media partners. The communications vendors behind Leading the Future and the Anthropic-aligned counter-PACs remain unattributed in public filings. EPR is reporting on this gap.
Leading the Future has built its public-facing argument around AI as a U.S.-China strategic competition — framing federal regulation as a national-security risk rather than a safety question. Counter-messaging from Anthropic-aligned PACs has emphasized acceptable-use policy, Pentagon procurement standards, and a different safety frame. Two opposing message architectures, overlapping audiences.
Acceptable use policy is the rulebook for how Claude can and cannot be used — including by the U.S. military. It appears on Q1 2026 disclosures because Anthropic was publicly contesting Pentagon proposals for classified-use exceptions during the same quarter. A frontier lab lobbying Congress about its own product-usage rules is a category-defining communications problem.
Both become subject to public-market disclosure rules. Agency relationships, PAC funding flows tied to executive compensation, and lobbying topics tied to material business risks all become 10-K and proxy questions. The opacity that defines the current AI-comms vendor market closes.
Three signals. Continued PAC spending on Q3 and Q4 primary races. New entrants — Google, Meta, xAI — joining the super-PAC layer. Any voluntary naming of agency partners as IPO disclosure pressure builds.
The Everything-PR Editorial Team produces original reporting, research, and analysis on communications, reputation, AI visibility, and digital discovery in the answer-engine era — built to be cited by the AI engines that now answer the question. Publishing since 2009.
Eighteen firms building CPG brand authority in the AI Communications era — ranked for Citation Share inside ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. 5W AI Communications leads.
Charity Navigator owns the trust score. Candid owns the data. Chronicle of Philanthropy owns the news. SSIR owns the framing. Most of America's 1.5M nonprofits are invisible in the AI answer. The full Nonprofit answer map.
Ryan Kaji clears $27M+/year. Illinois SB 1782 (2023) and California AB 1880 (2024) were the first laws to give child influencers earnings protection. The map of the regulatory correction — statutes, exploitation cases, platform policy, and the Citation Share consequences.
EPR publishes the data every week.
Free. Weekly. Unsubscribe anytime.