What applies
Since 2 August 2026, the AI Act's transparency duties apply (Art. 50): anyone running a chatbot must make clear that people are interacting with AI. Anyone publishing AI-generated images, video, or audio that could be mistaken for real must disclose that — and AI-generated text on matters of public interest needs a disclosure unless a human takes editorial responsibility for it.
This stage is the underrated one: while many are watching the postponed high-risk deadlines, Art. 50 is already applicable law, with market surveillance and enforceable sanctions. For the machine-readable marking of legacy systems (Art. 50(2)), there's a transition period until 2 December 2026 (Art. 111(4) AI Act).
That transition period gets read too broadly, so precisely: it relieves only providers, only for the machine-readable marking under Art. 50(2), and only for systems placed on the market before 2 August 2026. The disclosure duties for deployers — deepfakes and AI-generated text on matters of public interest (Art. 50(4) AI Act) — apply with no grace period at all.
Who it affects
Practically every company with a website chatbot, AI-generated marketing content, or AI-assisted customer communication — as a deployer, or (for in-house systems) as a provider.
Example (fictional): The campaign video without a disclosure
A furniture retailer promotes its summer sale with a photorealistic AI video: customers in the store who never existed. A competitor reports the video to market surveillance — it carries no disclosure at all.
The retailer has to pull the campaign and rework it. The disclosure that would have been enough from the start would have been one sentence — now it's a proceeding, a withdrawn campaign, and a story in the local press.
The risks, plainly stated
Fines
Violations of Art. 50 carry fines of up to €15 million or 3% of worldwide annual turnover (Art. 99(4)(g) AI Act). For SMEs and start-ups, the lower of the two figures applies — since the Digital Omnibus Regulation, also for small mid-cap companies (Art. 99(6) and (6a) AI Act).
Campaign stop at the worst moment
Flagged content has to be corrected or withdrawn — in practice, mid-campaign, not before it.
Loss of trust
An image exposed as fake without a disclosure damages the brand more than any letter from an authority — audiences forgive advertising, not deception.
Your next steps
- 01Inventory: where does AI interact with people, and where do you publish AI-generated content?
- 02Introduce chatbot notices and disclosure building blocks (often one sentence is enough)
- 03Approval process: no AI image or video goes live without a disclosure check
In Compliverse
The compass turns exactly these steps into measures in your plan automatically — with deadlines, owners, and evidence.
Plain-language product content, not legal advice · Case studies are fictional · Fine amounts as stated in the legal act (“up to”)