Definition & Explanation
The use of AI to evaluate or classify natural persons or groups based on their social behavior or personal characteristics, where that score leads to detrimental treatment. Social scoring is a prohibited AI practice under Article 5 EU AI Act because of its unacceptable risk to fundamental rights. The ban has applied since 2 February 2025 and covers both public authorities and private organizations.
Article 5(1)(c) EU AI Act prohibits AI systems that evaluate or classify persons or groups based on social behavior or known, inferred, or predicted personal characteristics, where the social score leads to (1) detrimental treatment in a social context unrelated to the context in which the data was originally collected, or (2) detrimental treatment that is unjustified or disproportionate to the behavior. Violations fall into the heaviest fine category: up to 35 million euros or 7% of global annual turnover, imposed by the supervisory authority.
The best-known example is the Chinese social credit system, but the prohibition is broader. Think of an insurer using social media behavior to set premiums for a product unrelated to that behavior, or a benefits agency assigning citizens a risk score based on neighborhood, language proficiency, and household composition. The Dutch SyRI system, which a court in The Hague struck down in 2020 for violating the ECHR, shows how close such systems can come to prohibited social scoring.
Social scoring is not just a government issue and not just "the Chinese system". The prohibition in the AI Act explicitly applies to private parties as well. At the same time, not every score is prohibited: a credit score based on relevant financial data falls outside the ban, although it is high-risk under Annex III. The test is whether data from one context is used for detrimental decisions in another, unrelated context, or whether the treatment is disproportionate.
