Anthropic just dropped another $20 million into the political arena. The move brings the company’s total commitment to $40 million this election cycle. And it lands at a moment when Silicon Valley’s divisions over artificial intelligence have hardened into open proxy warfare.
The recipient is Public First Action, a nonprofit that doesn’t have to reveal its donors. That structure gives it flexibility. It also draws scrutiny. The group channels resources toward advertising that spotlights candidates’ stances on AI safety. One early ad praised Republican Sen. Marsha Blackburn for her legislation on AI and child online safety. Blackburn is now eyeing Tennessee’s governorship.
Business Insider first reported the fresh donation and its context on July 22, 2026. The story laid out how Anthropic’s spending counters a rival effort backed by OpenAI President Greg Brockman, along with venture capitalists Marc Andreessen and Ben Horowitz. Their vehicle, Leading the Future, has pulled in more than $75 million. OpenAI has stressed that Brockman acts in his personal capacity there.
But the clash runs deeper than two super PACs. It reflects a fundamental split inside the AI industry. One side, anchored by Anthropic Chief Executive Dario Amodei, pushes for mandatory government testing of advanced models. The other warns that such rules could slow innovation to a crawl. Amodei made his position plain in an essay last month. He called for a new federal agency modeled on the Federal Aviation Administration. This body would require rigorous evaluations. It could even halt the release of a model if risks appeared too high.
The Trump administration and parts of the industry have pushed back against mandatory reviews. They argue the approach risks handing advantages to foreign competitors, particularly those in China. Yet Amodei and his allies see the alternative as worse. Unchecked development, they say, invites catastrophe.
Public First Action was cofounded by former Republican Congressman Chris Stewart of Utah and Democrat Brad Carson of Oklahoma. Their allied political action committees—Defending Our Values PAC and Jobs and Democracy PAC—back candidates from both parties. In June the nonprofit reported raising $80 million. By then it had already begun airing ads and supporting contenders up and down the ballot.
Anthropic insists the latest funds cannot directly influence any candidate’s election. They are meant for education on policy. Still, the nonprofit has routed money into messages that highlight specific politicians. The distinction matters legally. It also blurs in practice. Observers on X noted the tension immediately after the announcement. Some called it lobbying dressed in safety rhetoric. Others saw a necessary counterweight to pro-innovation money.
This isn’t Anthropic’s first foray. In May, Amodei personally gave $1 million to Public First, one of the allied super PACs. Earlier this year the company launched its own employee-funded PAC, AnthroPAC, according to The Hill. That vehicle allows voluntary contributions capped at $5,000 per person. It signals a broader, sustained push into politics.
The spending comes as AI-related groups have already poured tens of millions into the 2026 midterms. NPR reported in June that such super PACs had spent $43.3 million on congressional races by late spring. Fortune detailed a brutal $27 million battle in New York’s 12th congressional district. There, Anthropic-backed efforts supported one candidate while the OpenAI-aligned side attacked. The winner, Micah Lasher, promptly told both factions to stay away. He promised to regulate them anyway.
That race offered an early lesson. Heavy outside money doesn’t always buy gratitude. It can even backfire. Yet the flow of cash shows no sign of slowing. The SF Standard reported just five days ago that AI employees are donating faster and in greater sync than previous tech cohorts. Anthropic and OpenAI staffers funneled hundreds of thousands to candidates backing safety legislation. In California primaries, similar patterns emerged around gubernatorial and attorney general races.
Public First Action now sits at the center of the pro-safety camp. It explicitly opposes federal moves to preempt state-level AI rules without strong national safeguards. Reuters covered the latest $20 million pledge on July 22, quoting Anthropic’s statement that the donation aims to raise the urgency of the policy debate. As models like Claude Mythos advance, the company says, the need for measures to mitigate risks grows more pressing.
Brad Carson, who leads the group, told CNBC in early July that donations have come from employees at OpenAI, Google, DeepMind and X in addition to Anthropic’s corporate commitment. The nonprofit doesn’t disclose donors. Anthropic chose to make its own contribution public. Carson noted a noticeable uptick in AI-related bills and hearings on Capitol Hill. Passage this year looks unlikely given the short legislative calendar. Both parties, however, have flagged the issue as a long-term priority.
The OpenSecrets tracking organization has monitored the trend closely. Its data shows Anthropic quadrupled its federal lobbying outlays in 2025 to more than $3.1 million. The company walked away from a Pentagon contract rather than loosen safeguards on its Claude model. That decision, reported by OpenSecrets in March, underscored its safety-first posture even at the cost of revenue.
Critics see something else. Tech Oversight Project compiled a research book in June that tallied more than $165 million raised by AI-linked PACs and dark-money outfits. It accused the sector of trying to buy favorable rules while protecting profits. The report grouped Anthropic with other players but highlighted the safety-versus-innovation divide as a contest for control over AI’s future.
Washington Post reporting from May described super PACs intervening in primaries with ads on unrelated issues like immigration enforcement to boost favored candidates. One spot praised a Democrat for holding ICE accountable. The candidate also co-chairs a House commission on artificial intelligence. The tactic reveals how these groups treat elections as battlegrounds for broader influence.
Anthropic’s latest move intensifies that contest. It doubles down on a strategy of bipartisan outreach tied to concrete policy asks. Mandatory testing. Pre-release evaluations. Transparency requirements. Amodei has repeated these points in interviews and writings. He believes the United States must lead on responsible AI governance or risk losing the narrative to authoritarian regimes.
Yet the counterarguments carry weight in an election year. Innovation hawks contend that heavy regulation could drive talent and capital overseas. They point to export controls, compute restrictions and the rapid progress of Chinese labs. Brockman and his allies favor lighter-touch approaches that preserve speed. Their super PAC has spent aggressively to elect lawmakers who share that view.
The result is a proxy war playing out in House and Senate races across the country. Total AI-related independent expenditures have topped $50 million so far, per data aggregated by Transformer News. Pro-safety groups account for nearly $28 million of that. Pro-innovation efforts sit at $22 million. The gap is narrow. The stakes feel enormous.
Employees are voting with their checkbooks too. The SF Standard piece noted that 28 Anthropic and OpenAI staffers gave a combined $173,000 in a single day to one New York candidate last fall. Similar blitzes targeted California races. This individual giving often aligns with corporate priorities but operates independently. It adds another layer to the influence map.
Public reaction on X has been pointed. Some users mocked the safety framing as cover for market protection. Others praised Anthropic for refusing to sit on the sidelines. Chamath Palihapitiya and Suhail Doshi drew attention for criticizing the donation as token spend on politics rather than technology. Their comments spread quickly after the announcement.
Still, the spending continues. Anthropic’s statement tied the new funds directly to the development of advanced systems like Claude Mythos. Policymakers, the company argues, must grapple with risks now. Waiting invites regret.
That sense of urgency explains the scale. Forty million dollars buys a lot of ads, research and organizing. It also buys access. Carson has said the group’s work has already helped spark more serious conversations in Congress. Whether those talks produce law remains uncertain. The next few months of campaigning will test how much voters care about an issue that can sound abstract until it isn’t.
For industry insiders watching the midterms, Anthropic’s play offers a case study. It shows how a single startup can amplify its voice in national debates. It reveals the limits of purely technical competition when policy becomes a bottleneck. And it highlights the strange bedfellows that emerge when safety concerns cross party lines.
The proxy war won’t end in November. Winners on both sides will carry these fights into the next Congress. Amodei has signaled his commitment is long term. So has the opposing camp. What began as quiet lobbying has become a loud, expensive contest for the rules that will govern AI’s next decade.
Expect more announcements. More ads. More friction. The technology is moving fast. The politics are finally catching up.
WebProNews is an iEntry Publication