AI-Governance Claims, Checked Against Primary Sources
Why this report exists
Most claims about AI never get independently checked. They get repeated. In a field moving this fast, a "fact" that was true six months ago can be repealed, deferred, or superseded — and the decisions built on it inherit the error, silently.
This is the first in a standing series. We take claims circulating widely in the AI-governance market and check each against the primary source — the statute, the regulator, the official record. Nothing here is a hot take; every verdict is sourced. Two disciplines govern the report:
- We verify claims, not claimants. No company is named. We are not grading anyone's product or character — only whether a statement holds against the record.
- Precision over posture. The goal is not to "debunk." It is to state the exact position, including the nuance. An overcorrection is just a new error.
Verdict scale · VERIFIED / PARTIALLY VERIFIED / NOT VERIFIED
The European Commission's Digital Omnibus on AI (proposed Nov 19, 2025) deferred the high-risk compliance deadlines. Political agreement was reached May 6–7, 2026; the European Parliament endorsed it June 16, 2026; and the Council of the EU gave final approval on June 29, 2026. The deferred dates:
- Annex III (stand-alone high-risk systems): now December 2, 2027 — a 16-month deferral.
- Annex I (AI embedded in regulated products): now August 2, 2028.
The nuance most corrections miss: the deferral is targeted, not total. Article 50 transparency obligations remain effective August 2, 2026 — they were not moved. (Prohibited-practice rules have applied since Feb 2025; GPAI obligations since Aug 2025.)
"Everything is enforced Aug 2, 2026" is wrong. "Nothing happens Aug 2, 2026" is also wrong. High-risk obligations moved to Dec 2, 2027 / Aug 2, 2028; transparency obligations stay on Aug 2, 2026.
Colorado's original AI Act (SB 24-205, signed May 2024) never took effect as enacted. Its enforcement was stayed by a federal magistrate on April 27, 2026, and on May 14, 2026 Governor Polis signed SB 26-189, which repeals SB 24-205 and replaces it with a disclosure-and-rights framework for automated decision-making technology (ADMT), effective January 1, 2027.
SB 205 is no longer operative law. Any 2026 compliance plan, market map, or "why now" built on "Colorado SB 205" is built on a repealed statute. The live instrument is SB 26-189 (ADMT), effective Jan 1, 2027.
As of early-to-mid 2026, 23–24 states plus the District of Columbia have adopted the NAIC Model Bulletin on the Use of AI Systems by Insurers or substantially similar guidance — accurately described as "nearly half," not a majority. The operative 2026 development is the NAIC AI Systems Evaluation Tool, a 12-state pilot running January–September 2026 that brings insurer AI governance into market-conduct examinations.
"Nearly half the states (≈23–24 + DC)" is correct; "over half" is not. The examination tool — not the raw adoption count — is what changes insurer obligations in 2026.
The pattern — and why it matters
Three widely-repeated claims. Within roughly six months, all three moved: one deadline deferred, one law repealed, one count overstated. None of these shifts were hidden — they are in the public record. They persist in circulation because the claims were repeated, not re-checked.
That is the quiet risk under the loud AI-governance conversation. A compliance roadmap, a board memo, a market thesis, or an AI system's own "we're compliant" dashboard is only as current as the last time someone independent checked its inputs against the source. Self-report and repetition are not verification. This report is a small, public demonstration of the difference.
What this report is — and isn't
- It is an independent check of public claims against primary sources, issued at no charge.
- It is not legal advice, and not a compliance opinion for any organization. Confirm your own obligations with counsel and the current Official Journal / statute text.
- It names no company and evaluates no product. It grades statements, not people.
MarginSignal OS — independent verification of what deployed AI, and the claims around it, actually do.
Corrections to any finding in this report are welcome and will be issued publicly. That is the point.