The EU’s TikTok case targets the ranking signal, not merely harmful videos

The European Commission has preliminarily found that TikTok breached the Digital Services Act by failing to assess and mitigate the “rabbit hole” effect of its For You feed. The concern is not simply that individual videos about weight loss, self-harm, sadness or body image exist. It is that repeated viewing and engagement cause the recommender to supply progressively more of the same material, potentially worsening anxiety, low self-esteem, body dissatisfaction or existing eating disorders—especially among minors and vulnerable adults.

The proposed remedy is unusually specific. Brussels says TikTok should rely more heavily on users’ explicit statements about what they want, consider removing viewing time as a recommendation signal, make feed controls easier to find and use, and test the system regularly. The Commission’s investigation suggests that pressing “Not interested” may not be enough because inferred attention can carry more weight than an expressed preference.

That distinction reaches the core product. Watch time is attractive because it is continuous, abundant and behaviourally honest in a narrow sense: it records what held the screen. But attention is ambiguous. A person can linger because content is distressing, confusing or difficult to escape. Treating every extra second as positive preference can turn a prediction system into an amplifier precisely when the user’s welfare diverges from engagement.

TikTok rejects the findings as fundamentally inaccurate and says the Commission has distorted both its architecture and its safety work. The finding is preliminary, not a final judgment or fine. TikTok can inspect the evidence and respond; the European Board for Digital Services will also be consulted. If confirmed, a non-compliance decision could carry a penalty of up to 6% of worldwide annual turnover.

The case therefore sets up a consequential test for European platform regulation. Earlier enforcement focused on illegal content, transparency or addictive interface features. Here the regulator is asking whether the relative weighting of behavioural and explicit signals creates a systemic risk. If the theory survives TikTok’s defence, DSA compliance will no longer stop at adding safety tools around a recommender. It will constrain what the recommender is allowed to optimize.