Every AI-Sent Message Into the EU Now Needs a Disclosure. Most Channels Can't Physically Carry One

The AI Act's transparency rules became enforceable in August. A disclosure has to survive a 160-character text message, a language you don't speak, and a platform that strips formatting — here's how Kanvo makes that automatic, and refuses the send when it can't.

The notice has to survive the channel

There is no single sentence that works everywhere:

Saying it is not enough. Proving it is.

Here is the part most platforms have not thought about: after the enforcement date, "we have a disclosure policy" is not a defence. The question a regulator asks is what did this specific message say, and did it disclose?

The human in the loop is evidence too

Drafts are risk-scored before delivery. Anything over your threshold diverts to a named human approver instead of going out — and the approver's decision, with their identity and a timestamp, is itself signed and appended to the ledger.

What to do this week

Publish a policy version with your disclosure and risk settingsIssue a mandate to the API key your agent uses — scoped, expiring, revocableSend one test message through each channel you use, and check the disclosure badge in the inbox

Kanvo Team

Writes here about what Kanvo is building and why — the send path, the evidence it leaves, and the decisions behind both.

Scoped, expiring, revocable authority for AI agents as W3C Verifiable Credentials — plus a disclosure engine that enforces EU AI Act Article 50 per channel and per language.

Agent Mandates and the EU AI Act: Authority Your Agents Can Prove