From 2 August the AI Office and national authorities are enforcing the AI Act, and the transparency obligations apply: the chat says it's a machine, the deepfake carries a label, generated content carries a machine-readable mark. The Commission also published a list of over 180 organizations that have signed the transparency code.
- From 2 August, interactive AI systems must tell the user they're talking to a machine, and generated or altered content must carry a machine-readable mark.
- The Commission opened a complaints tool, a whistleblower channel, and a separate channel for complaints from downstream providers.
- Over 180 organizations have signed the code of practice on AI content transparency.
Some dates pass quietly. This one didn't. Since Sunday the AI Act has been enforced for real, and the transparency rules we talked about in July are now an obligation with an institution behind it.
What's new here isn't the rules - we knew those from July. What's new is the plumbing of enforcement. A complaint is filed with a form. A tip can also come from an employee inside. A complaint can also be filed by a company that builds on someone else's model and feels squeezed by the terms. And the list of code signatories is the other half of the mechanism - a public way to show you're one of the compliant ones, before anyone asks.
Here at home, enforcement now has an address. Until now, non-compliance was a matter of conscience. Since Sunday it's a matter of who points you out. And the most motivated auditor doesn't sit in Brussels - it's your competitor, one form away.
If AI talks to your customers or makes content for them, check off three things this week: a sentence in the chat, a label on the image, a machine mark in the file. Get it done before someone asks. Asking now has a form.