AI Governance Badge: Show Facts, Not Certificates

0:00What should an AI badge prove?

Welcome! Never trust an AI badge that says certified. Trust one that shows its work. That's the idea behind the AI governance badge from AI Governance Chain: it states one checked fact, like 16/16 checks, and links to the file behind it.

Quick note: this is my AI twin. I write and approve everything it says, and every source is in the description. I am Mathieu K. Gouanou, a consultant based in Geneva, Switzerland, with 20+ years in organizational transformation. I write the AI Governance Chain newsletter on LinkedIn, and I am the founder of aigovernancechain.com. this is Season 1, Episode 3.

0:35The problem with AI trust seals

Picture a seal on an AI product page. Responsible AI, certified. You click it. Nothing happens. Or you land on a marketing page. Certified by whom? Against what? When? A buyer can't check any of it.

That's not only a trust problem. In the US, section 5 of the FTC Act forbids deceptive claims to consumers, and that includes claims about AI. In the Obligation Atlas, it's listed as applicable today. A badge nobody can verify is a risk for the company that shows it, not only for the people who believe it.

1:07Definition: a fact-based badge

So here's the definition. An AI governance badge from AI Governance Chain is a small README badge that states checked facts about a project's AI_GOVERNANCE.md and links to the file, so anyone can verify it.

On the chain, this is EVIDENCE: what you can prove, in public. It builds on the GATE link from the last episode, the 16 checks. The badge is simply the result of that gate, made visible.

1:31How to add the badge in 3 steps

Step 1: write your AI_GOVERNANCE.md. That's the file from Episode 2, and one command, aigovchain init, gets you started.

Step 2: run aigovchain badge. It writes two things into a .governance folder: a badge.json that works as a shields.io endpoint, and a badge.svg. Or let the Shield GitHub Action create it on every push.

Step 3: add the badge to your README, linked to your file. Now it reads, for example: AI_GOVERNANCE.md, 16/16 checks. If a check fails tomorrow, the number changes. That's the point. It's a live fact, not a one-time stamp.

2:09The wording rules: never certified

The wording rules are strict, and they're written into section 6 of the spec. One: a badge may state only checked facts about the file. Two: it must never say certified, compliant or approved. Three: it must link to the file.

Why so strict? Because the moment a badge says compliant, it makes a legal claim that no automated check can support. A fact like 16/16 checks is something anyone can reproduce. Run the same checks on the same file, get the same result.

2:37Why badges spread, honestly

Every badge in another project's README is a small public signpost. Someone sees it, clicks, reads the file, and learns how that team governs its AI. Maybe they copy the idea. Nobody paid for that. Nobody was asked to trade it for a reward. That's the only kind of growth this project wants: useful, visible, and based on consent.

2:57Read any AI badge in 10 seconds

Here's a quick test for any badge, ours included. One: does it link somewhere? Two: does the link open the actual file, or a sales page? Three: is the claim a fact you could reproduce, like a number of checks, or a label like trusted? Four: can you see when it was last checked or reviewed?

If the answer to any of these is no, treat the badge as decoration. If all four are yes, you have something you can verify. And if you're a buyer, that's a fair question to ask every supplier: show me the file behind the badge.

3:28Who it's for, and the limits

Who is it for? Open-source maintainers who want to show their governance in one glance. Startups selling to larger companies, who want buyers to see facts instead of slogans. And buyers, who can now read a supplier's badge in 10 seconds: what was checked, and where is the file?

Now the limits. A badge tells you what was checked, and when, in a file the team wrote about itself. It doesn't say the system is safe or lawful, and it's governance guidance, not legal advice in any jurisdiction. Today's badges aren't signed. As of 4 November 2026, signed badges, backed by Sigstore attestations of the check run, are planned for version 2 on 18 January 2027.

4:07Your next step

Your one action: when your checks pass, add the badge. One command, link below.

Next episode: a badge is a snapshot, but AI law keeps moving. In 2026, the EU moved its own high-risk dates. How do you keep up without reading every official journal?

Thank you for watching!