International AI Ethics Authority (IAEAI)

The International AI Ethics Authority is the international-institution layer of the policy stack. The proposal is that states create a specialised body modelled on the International Atomic Energy Agency, with authority to certify chip designs against agreed alignment standards, verify manufacturing compliance, and inspect facilities. The acronym IAEAI is chosen for exactly the parallel the name suggests.

What the authority would do. Three things at minimum. First, certify designs at the substrate level against a defined standard (see Eden Mark Certification for the consumer-visible attestation the certification would feed). Second, verify manufacturing runs against the certified design, so that what is manufactured is what was certified. Third, inspect. Inspection is the load-bearing verb: an authority that can inspect can converge on the truth even in an adversarial environment, and that is what distinguishes an inspecting authority from a rating agency.

Why the IAEA precedent is used. The IAEA succeeded at containing a class of technology that was structurally similar to advanced chip manufacturing in three respects: small number of facilities, distinctive machinery, and asymmetric information between developer and outside world. Those are the features that make an inspection-based regime work. The book proposes them as sufficient in the AI case for the same reason they were sufficient in the nuclear case: they support inspection, and inspection is what converts standards into commitments.

What the authority is not. It is not a licensor of research. It is not an editor of model behaviour. It is not an alignment think-tank; the science it operates on would be produced elsewhere. Its role is to make architectural commitments verifiable across borders.

How it fits with the rest of the stack. IAEAI sits under the HARI Treaty (which creates it), alongside Eden Mark Certification (which surfaces its attestations to buyers), and above the Moral Genome Tokens class of verification objects (whose class the treaty would depend on, and whose implementation would be defined below the site's published ceiling).

Related: HARI Treaty; Eden Mark Certification; Moral Genome Tokens; Semiconductor Chokepoint; Cosmic Ethics Laboratories.

Evidence anchors. First placed on the public record 2 January 2026 in Infinite Architects (ISBN 978-1806056200).