KInIT and the Slovak Academy of Sciences advocate for cleared guidelines on red lines in the EU AI Act

Kempelen Institute of Intelligent Technologies (KInIT) and the Slovak Academy of Sciences submitted contributions to consultation on the European Union’s AI Act focusing on prohibited practices outlined in Article 5.

As red lines are part of our research activities under the project TECHNÉ, we highlighted concerns about widely used recommender systems, particularly on social media, that already engage in problematic behaviours affecting vulnerable individuals. For instance, these systems exacerbate financial losses for people with compulsive buying tendencies, recommend triggering content to individuals with PTSD or depression, and manipulate children and youth into unhealthy habits and addictions. The impact extends to individuals with eating disorders, ADHD, and other vulnerabilities, worsening both their mental and physical health.

Our input also addressed social scoring, which is not explicitly defined in the AI Act but aligns with behavioural analysis conducted by social media platforms. Concerns were raised about unjustified or disproportionate treatment of individuals through content personalisation and the formation of disinformation-driven filter bubbles. In the context of workplace and educational settings, KInIT and the Slovak Academy of Sciences emphasised the need for clearer guidelines on exceptions – particularly to prevent misuse of AI under the pretext of safety, as illustrated by frequent but unsubstantiated threats to institutions in Slovakia.

Additionally, we suggested refining the language in Article 5 (1) (g) regarding biometric categorisation systems. Specifically, we proposed adjusting the wording to improve methodological accuracy and avoid redundancy.

These contributions underline the importance of stronger safeguards for recommender systems, clearer exception criteria, and precise language to ensure the effectiveness of the regulation.