The $145 Million AI Election War: How Anthropic and OpenAI Are Buying America's Regulatory Future

The Most Expensive AI Fight Is Not in a Lab

On February 12, 2026, Anthropic announced a $20 million donation to Public First Action, a political action committee supporting candidates who favor AI safety regulation. The announcement was unremarkable by Washington standards — companies donate to PACs constantly. What made it significant was the target.

Public First Action exists to counter Leading the Future, a super PAC that has raised $125 million from OpenAI co-founder Greg Brockman, venture capital firm Andreessen Horowitz, and a coalition of Silicon Valley investors determined to elect candidates who will keep the government away from artificial intelligence.

Combined, these rival PACs have committed over $145 million to the 2026 midterm elections. The candidates they elect will determine whether the United States regulates AI like Europe — with mandatory safety assessments and transparency requirements — or continues the current approach of minimal federal oversight and active suppression of state-level regulation.

This is not a lobbying campaign. It is a war fought with campaign donations, attack ads, and competing visions of what responsible AI development looks like.

Follow the Money

Leading the Future: The $125 Million Machine

Leading the Future launched in August 2025 and entered 2026 with $70 million in cash on hand. The donor list reads like a directory of Silicon Valley's most powerful figures.

Greg Brockman, OpenAI's president and co-founder, and his wife Anna contributed $25 million — the largest individual AI PAC donation in history. Brockman's donation is particularly notable given OpenAI CEO Sam Altman's dramatic reversal on AI regulation (more on that below).

Marc Andreessen and Ben Horowitz, co-founders of the venture capital firm Andreessen Horowitz (a16z), each contributed $12.5 million personally. Their firm has invested billions in AI companies and published the "Techno- Optimist Manifesto" identifying regulation as a fundamental threat to technological progress.

Other donors include Joe Lonsdale (Palantir co-founder, $250,000 to affiliated PACs), Ron Conway (SV Angel founder, $500,000 to affiliated PACs), and Perplexity (AI search startup, $100,000).

The operation is sophisticated. Leading the Future is not a single PAC — it is a network:

• Leading the Future — the primary super PAC
• American Mission — a Republican-focused super PAC ($5 million from Leading the Future)
• Think Big — a Democratic-focused super PAC ($5 million from Leading the Future)
• Build American AI — an affiliated 501(c)(4) nonprofit "dark money" advocacy arm that can run issue ads mentioning specific candidates without disclosing donors

This dual-party structure mirrors the Fairshake crypto PAC playbook: fund candidates on both sides of the aisle, ensuring that AI deregulation is not a partisan issue but a bipartisan consensus backed by overwhelming financial force.

Public First Action: Anthropic's Counter

Anthropic's $20 million to Public First Action represents the company's first major political contribution beyond traditional lobbying — a significant escalation from a company that has historically engaged through policy papers and congressional testimony rather than campaign donations.

Public First Action plans to support 30 to 50 candidates from both parties in state and federal races. The group has already launched advertising campaigns for:

• Senator Marsha Blackburn — Tennessee gubernatorial race (Republican primary August 6, 2026), with ads focused on children's online safety
• Senator Pete Ricketts — Nebraska Senate reelection, with ads focused on AI chip export controls

Public First Action is seeking to raise $50 to $75 million total, but even at that ceiling, the pro-regulation side faces a roughly 2:1 funding disadvantage against the combined resources of Leading the Future and Meta's American Technology Excellence Project.

Meta's Third Front

Meta opened a third front in September 2025 with the American Technology Excellence Project, seeded with $20 million and focused specifically on state-level races. The PAC targets state legislators and candidates who oppose "onerous state-level AI and tech policy bills." Run by Republican veteran Brian Baker and Democratic firm Hilltop Public Solutions, the operation plans to spend "tens of millions" targeting state candidates across the country.

Meta's intervention reflects a specific concern: the company spent $19.7 million on federal lobbying in the first nine months of 2025 — a record — and views state-level AI regulation as a more immediate threat than federal legislation.

The Races That Matter

The theoretical debate over AI regulation is becoming tangible in specific congressional races where real candidates face real attack ads funded by AI industry money.

Texas 10th Congressional District

American Mission PAC (Leading the Future's Republican affiliate) is running ads supporting Chris Gober, a lawyer who served as chief counsel for Elon Musk's super PAC. Gober is running to replace retiring Representative Michael McCaul — who had supported restrictions on AI chip exports to China. The ad calls Gober a "Trump conservative" who supports "promoting American technology investment." Gober has endorsements from President Trump, House Speaker Mike Johnson, Governor Greg Abbott, and Senator Ted Cruz.

The race is a direct proxy for the AI regulation debate: McCaul supported chip export controls that frontier AI labs oppose. His replacement will likely vote the opposite way.

New York 12th Congressional District

Think Big PAC (Leading the Future's Democratic affiliate) spent $1 million on ads opposing Alex Bores, a Democratic state assemblymember running to replace retiring Representative Jerry Nadler. Bores co-sponsored and shepherded the RAISE Act — New York's AI safety law — into law.

The attack ad calls Bores a hypocrite for his past work at Palantir while now pushing for AI guardrails. Bores has made AI safety central to his campaign platform, and on February 12 — the same day Anthropic announced its donation — he released a comprehensive AI policy plan.

The Bores race is the most revealing proxy in the 2026 cycle. A Democratic AI safety champion is being attacked with millions of dollars from a PAC funded by Democratic-leaning tech billionaires. If Bores wins despite the spending, it signals that AI safety is a viable campaign issue. If he loses, it signals that financial force can neutralize safety advocacy even within the Democratic Party.

Sam Altman's Regulatory U-Turn

Understanding the current PAC battle requires understanding how OpenAI's position on regulation reversed completely in two years.

May 16, 2023: The Safety Champion

When Sam Altman testified before the Senate Judiciary Committee in May 2023, he positioned himself as AI's most responsible steward. He told senators: "My worst fears are that we cause, we the field, the technology, the industry, cause significant harm to the world. I think if this technology goes wrong, it can go quite wrong."

He proposed a three-point regulatory plan: a federal agency to grant licenses to AI models above a capability threshold, safety standards for high-capability models, and required independent audits. He mentioned AI safety dozens of times during his testimony.

May 8, 2025: The Innovation Champion

When Altman returned to the Senate two years later, the message was unrecognizable. He warned that requiring government approval to release AI software would be "disastrous" for America's competitive position. He urged lawmakers against regulations that could "slow down" the US in the AI race against China. He supported a "light touch" federal approach and a blanket preemption of state regulations across all 50 states.

References to AI safety, which dominated his 2023 testimony, were notably absent.

The SB 1047 Contradiction

The reversal became explicit during California's SB 1047 debate in August 2024. OpenAI's Chief Strategy Officer Jason Kwon sent an opposition letter arguing the bill would "threaten California's growth, slow the pace of innovation, and lead California's world-class engineers and entrepreneurs to leave the state."

Fifteen current OpenAI employees signed an open letter supporting SB 1047, directly contradicting their company. Former employees William Saunders and Daniel Kokotajlo wrote that they resigned because they "lost trust" in OpenAI's commitment to safety.

The trajectory is clear: OpenAI evolved from calling for federal licensing in 2023 to funding a $125 million PAC opposing regulation in 2026. Whether this reflects genuine belief that regulation would harm innovation or strategic positioning to protect market dominance depends on whom you ask.

The Philosophical Divide

The PAC battle is the financial expression of a genuine intellectual disagreement about how to govern the most powerful technology since nuclear energy.

The Safety-First Argument

Anthropic's CEO Dario Amodei published a 20,000-word essay titled "The Adolescence of Technology" on January 27, 2026, laying out the safety-first case with unusual directness. He warned that "humanity is about to be handed almost unimaginable power, and it is deeply unclear whether our social, political, and technological systems possess the maturity to wield it." He stated that "we are considerably closer to real danger in 2026 than we were in 2023."

Amodei identified five categories of existential risk: autonomous misalignment, biological misuse, authoritarian consolidation, massive economic disruption, and concentrated power. He proposed transparency laws, export controls, mandatory disclosures about model behavior, and incremental regulation. He cited California's SB 53 and New York's RAISE Act as templates.

Importantly, Amodei "repeatedly stresses that regulations should reduce hurdles imposed on smaller, nascent AI companies that are not operating on the frontier of AI." This directly addresses the regulatory capture accusation — the argument that Anthropic supports regulation because it can afford compliance while competitors cannot.

The Innovation-First Argument

Andreessen Horowitz's "Techno-Optimist Manifesto" identifies regulation as a fundamental threat to human progress. The manifesto lists "statism, authoritarianism, collectivism, central planning, socialism" as enemies and argues that regulatory regimes are built on "crude industrial revolution-era technology models" that risk "harming more people than they save with 'safety' measures."

Leading the Future's stated goal is to support candidates who favor AI innovation and oppose candidates who "believe AI is moving too quickly and want to rein in the technology." The underlying premise is that the United States cannot maintain its AI leadership if companies need government permission to deploy new models, and that China's lack of similar constraints creates a competitive disadvantage that outweighs safety concerns.

David Sacks, serving as the White House AI and crypto czar, accused Anthropic of "running a sophisticated regulatory capture strategy based on fear-mongering." He posted on X that Anthropic is "principally responsible for the state regulatory frenzy that is damaging the startup ecosystem." He further claimed "the real issue is not research but rather Anthropic's agenda to backdoor Woke AI and other AI regulations through Blue states like California."

What Neither Side Acknowledges

The safety-first camp underplays the genuine burden that compliance places on smaller companies. Even well-designed regulation creates costs that large companies absorb more easily than startups. New York's RAISE Act initially targeted developers spending over $100 million on compute — a threshold that only a handful of companies reach — but was later amended to $500 million annual revenue, acknowledging the startup impact concern.

The innovation-first camp underplays the genuine risks. When Microsoft's own research confirms that 70 percent of security vulnerabilities stem from memory safety issues, and AI systems demonstrate increasingly autonomous capabilities, dismissing safety concerns as "fear-mongering" requires ignoring evidence that the companies building these systems take seriously internally — including OpenAI, which maintains its own safety team and publishes system cards for major model releases.

The State Regulation Battlefield

While the PAC battle targets federal races, the most consequential AI regulation is happening at the state level — which is precisely why both sides are spending so aggressively.

The Failed Federal Moratorium

In May 2025, the House passed a 10-year moratorium on state AI regulation enforcement as part of the reconciliation "Big Beautiful Bill." Senator Ted Cruz attempted to insert it into the Senate version. The provision was watered down to a five-year "temporary pause" and then, in a stunning rebuke, killed by a 99-1 Senate vote. President Trump signed the bill without any AI regulation restrictions on July 4, 2025.

The 99-1 vote demonstrated overwhelming bipartisan opposition to federal preemption of state AI authority. Both parties' senators recognized that their states wanted the ability to regulate AI — a signal that state-level regulation will continue regardless of federal inaction.

Where State Regulation Stands

California SB 53 (effective January 1, 2026): Requires frontier AI developers to publish safety testing protocols publicly. A transparency-focused successor to the vetoed SB 1047 that shifts from strict safety requirements to disclosure mandates.

New York RAISE Act (signed December 19, 2025): Requires large AI developers with over $500 million annual revenue to create and publish safety frameworks, report incidents within 72 hours, and face penalties of up to $1 million per violation ($3 million for subsequent violations).

Colorado AI Act (delayed to June 30, 2026): Focuses on high-risk AI systems in education, employment, financial services, housing, healthcare, and legal services. The second state to enact major AI consumer protection law after Utah.

Illinois (effective January 1, 2026): AI disclosure requirement for employers using AI in employment decisions.

This patchwork of state regulation is exactly what Leading the Future and Meta's PAC are spending to prevent, and exactly what Public First Action is spending to protect.

The EU Comparison

While America's AI governance is being shaped by competing PACs, the European Union is implementing the most comprehensive AI regulatory framework in history.

The EU AI Act entered into force on August 1, 2024, with the majority of provisions becoming applicable on August 2, 2026 — less than six months from now. The act uses a risk-based categorization system (unacceptable, high, limited, and minimal risk) with mandatory bans on certain AI uses and strong obligations based on risk tier. Penalties reach up to seven percent of global annual turnover for prohibited AI violations. Critically, the act has extraterritorial application — affecting any company serving EU users.

The divergence between US and EU approaches is now fundamental. One analysis characterized the philosophies as "fundamentally incompatible governance models." The EU emphasizes comprehensive, rights-based and risk-based legislation. The US deregulatory approach prioritizes technological dominance over safety constraints.

For AI companies operating globally — which includes every major frontier lab — the EU framework creates a de facto baseline that must be met regardless of US policy. The question the PAC battle is really answering is not whether AI will be regulated at all, but whether the US will develop its own regulatory framework or simply defer to Europe's.

The Lobbying Escalation

PAC spending is only the most visible component of AI industry political engagement. Lobbying expenditures tell an equally important story of escalation.

Anthropic's lobbying spending trajectory illustrates the speed of the shift: $210,000 in 2023, $720,000 in 2024, and $3.13 million in 2025 — a 511 percent increase over the period. Monthly lobbying expenditures rose from $180,000 per month at the end of Q1 2024 to $1.1 million per month by Q4 2025.

OpenAI's lobbying spending hit $2.1 million in 2025, a 68 percent increase from $1.8 million in 2024 — significant growth but starting from a much lower base than Meta's $19.7 million record or Alphabet's $12.2 million.

Total AI-related lobbying across all companies is now measured in the hundreds of millions when including trade associations, consulting firms, and corporate government affairs operations beyond direct lobbying registrations.

The Fairshake Precedent

The AI PAC battle follows a template established by the cryptocurrency industry. The Fairshake PAC and its two aligned groups spent a combined $290 million in the 2024 election cycle. Nearly half of all corporate dollars that flowed into the 2024 election came from the crypto industry. Fairshake aided more than 50 candidates' wins, including Ohio Republican Bernie Moreno who defeated Senator Sherrod Brown, and successfully opposed Democratic Representative Katie Porter's California Senate bid.

For the 2026 cycle, Fairshake and affiliate PACs have more than $193 million in cash on hand. Major backers include Coinbase (which gave $75 million in 2024 and committed another $25 million for 2026) and Ripple Labs.

Leading the Future explicitly modeled itself on Fairshake's playbook. The lesson crypto PACs demonstrated is that concentrated financial force can decisively influence election outcomes, even in races where the PAC's preferred candidate is not the natural favorite. If $290 million could reshape crypto policy, the AI industry's current $145 million commitment is likely a floor, not a ceiling.

What It Means for AI Startups

The debate's impact on startups is more nuanced than either side acknowledges.

State-level AI laws are already reshaping vendor contracting practices. Even companies below statutory thresholds face downstream compliance expectations through AI-specific contract addenda and third-party risk requirements. Insurers have begun introducing "AI Security Riders" requiring documented evidence of adversarial red-teaming, model-level risk assessments, and specialized safeguards.

If the anti-regulation camp wins decisively, startups avoid compliance costs but lose the transparency requirements that could differentiate safety- conscious companies from competitors cutting corners. If the pro-regulation camp wins, compliance costs increase but the regulatory framework could also create barriers to entry that protect established players.

The honest answer is that no regulatory outcome is unambiguously good for startups. But a fragmented state-by-state patchwork — which is what happens if the federal government does nothing — may be the worst outcome for everyone, creating 50 different compliance regimes without the clarity of either a national standard or a national exemption.

The $145 Million Question

The amount of money being spent on AI regulation is itself a signal about the stakes. Companies do not commit nine-figure sums to political campaigns unless the regulatory outcome directly affects their business model.

For OpenAI, valued at $157 billion, the difference between a regulated and unregulated AI market is measured in tens of billions of dollars. For Anthropic, valued at $61 billion, the same calculation applies — but with the added dynamic that the company believes regulation actually benefits its approach to development. For a16z, whose portfolio includes dozens of AI companies, the regulatory environment determines whether their investments generate returns or face compliance-driven margin compression.

The 2026 midterms will not resolve the AI regulation question permanently. But they will determine the composition of the Congress that will write — or decline to write — the first comprehensive federal AI legislation. They will determine which state legislators face financial consequences for supporting AI safety bills. And they will establish whether AI safety is a winning or losing political issue.

The $145 million being spent is not an investment in candidates. It is an investment in the regulatory framework that will govern the most powerful technology in a generation. The candidates are merely the mechanism through which that investment delivers returns.

Whether those returns benefit the public or merely the companies writing the checks depends entirely on which checks are larger. As of February 2026, the anti-regulation side has a substantial lead.