EU AI Act: distinguishing “manipulative” AI from standard adtech

AI-powered adtech = prohibited AI? Interesting perspective, though I have my doubts as to whether the typical adtech-related AI systems we encounter in Europe would be classified as “manipulative” (let alone harmful) under the EU AI Act – once we look at how things work in practice. After all, manipulation and influence are not the same thing, and as Dr. Markus Wünschelbaum admits, “regular advertising is also playing with our behaviour and trying to influence us”.

Are we really in a situation with “the objective, or the effect, of materially distorting the behaviour of that person or a person belonging to that group in a manner that causes or is reasonably likely to cause that person or another person significant harm”, to quote the AI Act itself? (see also the Commission’s new Guidelines on prohibited AI practices)

Profile-based advertising is vilified yet misunderstood (through perpetuation of myths or excesses that are not tolerated anyway in the EU), and it’s critical to look at how things work in reality. Online advertising in general is facing a number of challenges today, and the EU AI Act is just one of them (next to GDPR and ePrivacy enforcement – and again misunderstandings, notably re pseudonymisation).

Markus suggests that increased scrutiny is coming and that there will be verifications of whether adtech practices are prohibited AI practices. Authorities (like the Hamburg DPA for which Markus works) are not the only ones to look at, as civil lawsuits are also a possibility.

If you’re in the adtech space or in any other area on the edge of the scope of the AI Act’s prohibitions, you really need to consider how you document what you do – and how to limit the risk of a challenge (by authorities or campaigners) being successful. Reach out – let’s chat!

Podcast interview of Markus by
Sergio Maldonado as part of his Masters of Privacy series: https://lnkd.in/e6vtjJyv

Data protection DarkPatterns