Polish ad industry sorts out who labels AI content as AI Act transparency rules bite

The transparency obligations of the EU AI Act have applied since August 2, and the Polish advertising market has spent the month working out what they mean in practice. The emerging consensus, laid out in a guide from IAB Polska’s AI and law task force and in a Wirtualne Media round-up of agency and ministry views published August 28, is narrower than many marketers feared: using an AI tool somewhere in production does not by itself trigger an on-screen disclosure.

What does trigger it is a deepfake as defined in Article 3(60): AI-generated or manipulated image, audio or video that resembles real people, objects, places or events closely enough to pass as authentic. Jowita Bieda of the Ministry of Digital Affairs told Wirtualne Media that the mere use of AI in producing an ad does not decide whether a label is required. Beata Michalik of Labcon summed up the rule as labelling only what pretends to be real, and Paweł Mucha of dentsu Poland noted that color grading and similar touch-ups fall outside it.

The IAB Polska guide, written by lawyers from TKP, Publicis Groupe and WPP Media, distinguishes two layers. Providers of AI systems must embed machine-readable marking in outputs. Deployers, which in advertising usually means the agencies or in-house teams operating the tools, must add a visible, clearly distinguishable notice at first exposure whenever the result is a deepfake. Metadata, a footnote in terms of service or a string in a URL does not count, and text is largely exempt unless it informs the public on matters such as health or safety.

Why it matters

Responsibility follows control, not job titles. Whoever decides how the AI is used, whether an agency, a production house or the brand’s own team, is the deployer and carries the duty, which is why the guide urges advertisers to fix roles and labelling standards in contracts. Fines can reach €15 million or 3% of global annual turnover. Marta Botiuk-Filip of the Czyżewscy law firm warned this week that a small gray line at the bottom of a creative does not amount to informing the audience, and that many companies still have no internal rules on which AI tools staff may use.

What’s next

Enforcement in Poland is still being assembled. The new Commission for the Development and Security of Artificial Intelligence, which will draw on UOKiK, UKE, KNF and KRRiT, is expected to have its chair chosen in October and to start work in November, according to Wirtualne Media, and AI systems already on the market have until December 2 to meet the technical marking requirement. Until the regulator issues its own guidance, the IAB document is the closest thing Polish advertisers have to a rulebook, and a written internal procedure on when a creative needs a label is the sensible minimum.