The EU writes the labeling rules in September, and deployers should be in the room
Signatories to the transparency Code are being invited into implementation working groups. That is where the practical labeling expectations for the 152 deployer signatories actually get decided.
Source: European Commission ↗
- EU AI Act
- AI governance
- Compliance
- Content provenance
The European Commission is standing up implementation task forces on the Code of Practice for Transparency of AI-Generated Content in September, and is inviting its roughly 190 signatories into working groups on shared practice and implementation standards.
The text is settled. The practice is not.
The Code names the practice, and the AI Act's transparency rules underneath make it law. Neither tells you what a compliant label looks like on a product page, or inside a support macro that fires four thousand times a day. That gets settled in the implementation work, and September is when it starts. The 152 companies that signed as deployers will be shaping the answer they then have to live with.
Everyone else inherits that answer without a seat. Standards work has always run this way. It is workable, provided somebody at your company is watching it happen.
What I'd do about it
Do not wait for the working groups to report before writing anything. The disclosure-policy work is small and specific, and the watermark piece walks through it: one written policy, a record of what was AI-drafted, language in the MSA. A policy you already have only needs tightening when the standards land.
Then name one person to track what the working groups produce. In most mid-market companies nobody owns this yet. Legal reads the Code and stops at the legal text; marketing reads the label requirement and misses the scope. Somebody has to hold both halves, and it is worth deciding who before the first customer asks. The published diagnostic walkthrough covers the inventory step you will need either way.