CTS-1: MAI-1 Conformance Test Suite — Binary Pass/Fail Enforcement for AI Attestation Infrastructure

CTS-1 is the binary pass/fail conformance test suite for the Model Attestation Interface (MAI-1), the canonical attestation protocol for foundation model governance. This public specification defines the complete compliance surface that any MAI-1 implementation must satisfy — 45 derived test assertions organized across a four-level test pyramid (structural, cryptographic, semantic, behavioral), three conformance tiers (MAI-C0 for research, MAI-C1 for commercial deployment, MAI-C2 for regulated and insured systems), and a 250+ vector negative test library covering malformed artifacts, cryptographic attacks, temporal exploitation, semantic consistency violations, behavioral manipulation, and grammar-based fuzzing.CTS-1 synthesizes conformance methodology from four established security certification traditions: Common Criteria (ISO/IEC 15408/18045) for tiered assurance and protection profiles, FIPS 140-3 for derived test requirements and binary precision, the OpenID Foundation conformance program for self-certification economics and operational agility, and the IETF RATS working group for protocol-level attestation interoperability. The resulting conformance model applies strictly binary pass/fail semantics — a system either passes all applicable assertions at its claimed conformance level or it fails. There is no partial compliance, no grading curve, and no interpretive discretion.The assertion registry decomposes every normative MUST and SHALL statement in MAI-1 §6–§9 into testable requirements with unambiguous pass/fail criteria. Each assertion specifies: the MAI-1 source requirement, the conformance levels at which the assertion applies, the test pyramid level (structural, cryptographic, semantic, or behavioral), and a deterministic pass/fail criterion. Statistical assertions for continuous invariant measurements employ the Truncated Sequential Probability Ratio Test (TSPRT) with controlled Type I and Type II error rates of 0.01 and invariant-specific distribution families.CTS-1 is Document 26 of the Auburn Governance Stack (AGS-1), positioned in the Enforcement Layer. It is to the Auburn stack what RFC 793 (TCP) is to the internet protocol suite: the reliability and enforcement mechanism that transforms a normative specification into a deployable, verifiable, and enforceable standard. Once published, any procurement officer can paste the assertion requirements into an RFP, any insurer can require test passage as a policy condition, and any regulator can reference the test suite in enforcement guidance.The document acknowledges fundamental limitations transparently: CTS-1 certifies process compliance, not behavioral safety. Rice's theorem precludes behavioral guarantees from finite test suites. Measurement truthfulness assurance increases across conformance levels (self-certification at C0, TEE-rooted signing at C1, adversarial behavioral testing at C2) but absolute truthfulness guarantees are beyond the scope of any conformance regime. CTS-1 provides the precondition for accountability — without a public, deterministic, reproducible conformance standard, AI governance remains aspiration rather than infrastructure.