Political Economy · Artificial Intelligence
The Danger Is Not an AI Monopoly. It Is a Government-Protected AI Oligopoly.
There is a difference between winning in the free enterprise system and using Washington to decide who else is allowed to compete in it.
There is an important distinction between winning in the free enterprise system and using the government to prevent others from competing in it. The first is how America built the most dynamic economy in history. The second is how mature industries protect themselves from the next generation of challengers.
Anthropic’s approach to AI safety deserves examination through the lens of regulatory capture. When companies with billions of dollars in capital help write the rules for their own industry, those rules can become barriers to entry for everyone else.
The incumbent can afford compliance. The entrepreneur cannot.
And Washington gets to call it consumer protection.
The Evidence Deserves Congressional Scrutiny
Anthropic’s own policy proposals and disclosures provide the starting point.
- Government gatekeeping. In his June 10 essay, “Policy on the AI Exponential,” CEO Dario Amodei called for FAA-style federal authority over frontier models, arguing their release “should be blocked or reversed as a threat to public safety” if they fail safety standards. — Axios
- Lobbying. Anthropic spent more than $3.5 million lobbying Washington in the first half of 2026, already exceeding its $3.1 million total for all of 2025. Its second quarter alone, nearly $2 million, was its largest since it began lobbying in March 2024. — Axios; Issue One
- Compliance architecture. The proposed framework calls for mandatory third-party testing of models trained above 1025 FLOPs or built by companies with more than $500 million in AI revenue, across cybersecurity, biological, automated-research and loss-of-control risk categories. — Anthropic policy roadmap, June 10, 2026
- Open-weight competition. OpenAI and Anthropic, fierce commercial rivals, aligned in Washington this summer in warning policymakers about powerful open-weight models, led by Chinese releases. Anthropic declined to sign the Nvidia-led industry letter defending open weights, drawing accusations that it favors restrictions on cheaper competitors. — Axios
Even more interesting is Anthropic’s IPO prospectus. The S-1 registration statement filed October 26, and amended ahead of the company’s November 18 pricing, tells investors plainly that its policy advocacy and executives’ public statements have drawn criticism from influential political figures, and that it intends to keep advocating. A company asking public markets for tens of billions of dollars is disclosing, in a federal filing, that shaping government policy is part of how it operates.
The Regulatory Capture Cycle
- Accumulate billions in private capital
- Establish technological dominance
- Emphasize catastrophic industry risks
- Lobby Washington for mandatory oversight
- Raise the cost of competitive entry
- Entrench incumbent market power
Illustrative mechanism of regulatory capture, not a finding of unlawful conduct by Anthropic.
Why This Could Be Bad for America
Imagine if IBM had persuaded Congress in 1975 that independent computer development presented an unacceptable national security risk.
Would Apple have survived? Microsoft? Dell?
Imagine if the established telephone monopoly had been handed regulatory authority over the emerging internet.
Anthropic will answer that its proposed thresholds, 1025 FLOPs and $500 million in AI revenue, exempt the garage startup entirely. Today, they do. But compute costs fall every year, and thresholds written for the frontier have a way of becoming the floor. More importantly, regulatory regimes are not static. Once a federal licensing apparatus exists, its scope is set by the agency, the incumbents who staff its advisory boards and the lobbyists who fund its champions, not by the startups it was promised to spare.
The next great American AI company may currently consist of three engineers working in a garage. Its greatest obstacle should be technological excellence, not obtaining Washington’s permission to compete against Anthropic.
The National Security Contradiction
We could conceivably restrict thousands of American developers in the name of competing with China while China continues expanding its own AI ecosystem, unburdened by the rules we write for ourselves.
We would be voluntarily narrowing America’s intellectual and entrepreneurial capacity.
There are legitimate cybersecurity and biological risks associated with advanced AI. Anthropic’s safety concerns should not be dismissed. To his credit, Amodei has said publicly that he has never advocated a ban on open-weight models, conceding that such a ban would shield U.S. AI companies from competition, “but that has never been my goal.” And Anthropic has itself been on the receiving end of federal power this year, from a supply-chain risk designation at the Pentagon to Commerce Department export restrictions on its own models.
Stated intentions, however, are not a governance structure. Neither Anthropic’s sincerity nor its scars should entitle Anthropic—or OpenAI, Google or Microsoft—to help determine who may compete.
Regulatory capture is a credible risk here, not yet proof of a deliberate monopoly conspiracy. That is precisely why Congress should scrutinize it now, before the rules are written, rather than after.
The WFPX Editorial Angle
The New Robber Barons Want a Federal License to Think
Yesterday’s industrial monopolists sought control of America’s railroads, oil and steel. Tomorrow’s may seek something considerably more valuable: control over America’s access to intelligence itself.
The Closing Argument
America did not become the world’s greatest economic power by asking established corporations for permission to innovate. We became that power by allowing entrepreneurs to challenge them.
Artificial intelligence must not become the exception.
Protect America from the misuse of artificial intelligence. But protect Americans from the monopolization of intelligence itself.