The AI Rating Bureau
Standardized rating once protected bureau insurers from independent pricing. AI regulation could now protect closed-weight AI labs from independent intelligence.
Concept and art direction: Onur Gungor / AllegoryOS. Illustration created with AI assistance.
In the 1920s and 1930s, two major bureau-independent insurers emerged in the United States. They developed more granular pricing and independent rating systems at a time when much of the industry relied on cartel-like rating bureaus.
Bureau pricing standardized rates across participating insurers. It also sheltered weaker operators and blunted the advantage of companies that understood their risks better. That two companies, State Farm and Allstate, did not simply accept bureau pricing as the only possible price.
These two new entrants developed their own actuarial models, classifications, and methods of selecting risk. In documented cases, their rates were materially below bureau rates for particular markets and risk classes. These companies used their mathematical and operating advantages to attract lower-risk, more profitable customers. That placed pressure on bureau insurers because the independents selected a more favourable risk mix.
The incumbents fought independent pricing through the regulatory system. The regulatory structure changed sharply in 1944, when the Supreme Court ruled that interstate insurance was subject to federal antitrust law. In 1945, the McCarran-Ferguson Act restored primary insurance regulation to the states and granted a conditional antitrust exemption. By 1947, most states had enacted rating laws that licensed bureaus and subjected rates to state oversight.
Independent insurers then spent years challenging rules that restricted their ability to charge less. In 1966, a federal appeals court described North Carolina’s prohibition on downward deviations as effectively preventing competition through lower premiums.
Regulatory fear has long been an effective anti-competitive shield. Today’s “AI Bureau” is using the same strategy against open-source and open-weight competitors.
The most powerful closed-weight AI labs repeatedly warn that advanced models could help create bioweapons, escape human control, eliminate white-collar employment, or create a permanent underclass. Dario Amodei has warned that AI could eliminate half of all entry-level white-collar jobs and push unemployment to between 10% and 20% within one to five years.
Behind every machine action is a chain of human decisions. A machine does not independently decide to build a bioweapon, replace a workforce, or create an underclass. People build and operate systems that make those outcomes possible.
On September 21, 2026, US Treasury Secretary Scott Bessent said responsibility for the OpenAI agents’ breach of Hugging Face belonged to OpenAI’s management. “Humans are responsible, not the AI,” he said.
Engineering is the discipline of asking what could go wrong and designing against it. If a builder gives a system unsafe authority, ignores a foreseeable failure, or sacrifices safety to protect profit and competitiveness, that remains a human decision.
Government does, however, have a legitimate role: to regulate demonstrated risks impartially, establish responsibility, and hold the people and companies exercising control accountable. It should not allow incumbent fear to determine who is permitted to compete.
Standardized rating once protected bureau insurers from independent pricing. AI regulation could now protect closed-weight AI labs from independent intelligence.
The Turkish poet and writer Mehmet Âkif Ersoy summarized this situation in “Kıssadan Hisse,” from Safahat:
“They define history as mere ‘repetition’; Had anyone ever learned a lesson from it, would it ever repeat itself?”