Genesis from Money Already at Risk
Lloyd's Register (1760) shows the easiest correction mechanism to build: one a self-interested counterparty would build anyway, because their own capital is exposed to hidden quality.
What decision changes?
Look for a counterparty whose own money is on the line and a signal cheap enough to check that faking it costs more than complying honestly — this genesis route needs no catastrophe and no prior political consensus.
The easiest safety institution to build is one that a self-interested party would build anyway. In eighteenth-century London, marine underwriters faced a problem that will sound familiar to anyone thinking about AI oversight today: they had to price the risk of something whose true condition they could not see. A ship’s owner knew whether the hull was rotten; the insurer did not. So in 1760 the underwriters organized their own answer — Lloyd’s Register, an independent inspection that surveyed ships and published a rating. A shipowner with an unclassified vessel found insurance expensive or unavailable, which made the inspection effectively mandatory without any law requiring it.
What makes this case remarkable is what it did not need. No state mandated it: the Register predates government maritime safety regulation by more than a century. And no disaster founded it: ordinary, continuous underwriting losses did the work, because every wreck was informative to the insurer whether or not it made the newspapers. The correction signal never depended on a tragedy crossing a public-attention threshold — a sharp contrast with drug regulation or aviation, which had to be bought with body counts.
The pattern recurs wherever someone’s own capital is exposed to another actor’s hidden quality. Financial audit began as a service investors demanded from joint-stock companies before any state required it. Today, cyber-insurance is repeating the pattern for computer security: insurers price ransomware exposure and require specific technical controls as a condition of coverage, effectively regulating a domain legislatures have struggled to reach.
For AI governance, this is the genesis route that asks the least of politics. It requires only two ingredients: an actor whose own money — not the public’s — is on the line, and a quality signal cheap enough to verify that faking it costs more than honestly meeting it. Deployment-liability insurance for AI systems, or compute providers bearing some liability for what runs on their infrastructure, are candidate levers of exactly this kind. They need no prior political consensus and no founding catastrophe to start functioning — only an insurer with the right incentives and something checkable to inspect. Given that the catastrophe-driven route may be unavailable for AI (see the catastrophe-ratchet card), this quiet, self-interested route deserves more attention than its unglamorous history suggests.
One of eleven historical case studies in Institutional Genesis, Memory, and Decay — see the overview for the full life-cycle map, or read the complete appendix.
What would count as evidence?
Marine underwriting losses informed classification ratings continuously, without needing a dramatic wreck to cross a public-attention threshold; the modern analogue is cyber-insurance pricing ransomware exposure and requiring specific technical controls.