Alongside the enforcement announcement, the Commission published the first signatory list for the Code of Practice on Transparency of AI-generated Content. Two implementation task forces launch in September 2026.
The arithmetic trap
Section 1, for providers, has 83 signatories. Section 2, for deployers, has 152. Those sum to 235, and that number is wrong: organisations can sign both sections, so the Commission reports about 190 distinct organisations. The Commission is not being consistent either — one of its two pages published the same day says "about 190" and the other says "over 180". Attribute the figure rather than asserting it.
Who is on which list
Section 1 includes Anthropic, Google, Meta, Microsoft, OpenAI, Mistral, Cohere, Aleph Alpha, Black Forest Labs and Synthesia. Section 2, the deployers, is a different kind of company entirely: Getty Images, Lenovo, Lufthansa, Bulgari and Iberdrola.
Signing is not the obligation
The Code is voluntary. It is a presumption-of-conformity route — a way to demonstrate compliance with Article 50, not the source of the duty. A company that signs nothing is under exactly the same legal obligation. Reporting that 190 companies have agreed to label AI content inverts the relationship: the labelling requirement is the law, and the Code is a way of showing your working.
Count is a weak proxy for coverage
Roughly half the signatories are small or recently founded companies, and the Commission notes the asymmetry itself: the legal duty falls on providers, so Section 2 deployers are volunteering beyond what the law requires of them. A headcount of signatures says little about how much of the European market is covered.
The office that has to run it
Two implementation task forces launch in September, and the AI Office is staffing up to meet them — 38 additional staff in Brussels, according to reporting the same day, together with a whistleblower channel aimed at technology workers and a confidential compliance-reporting tool. Both tools are an admission of the underlying problem: a regulator cannot inspect a model's training or marking practices from outside, so it is building routes for people on the inside to tell it what is happening.