Social Scoring by Public and Private Bodies
The third prohibited category covers what the EU calls social scoring. This is an AI system that assigns or predicts a score or rating to a person based on their social behaviour, character, values, or predicted future behaviour, and then uses that score or rating to restrict or deny the person access to services or opportunities that others can access.
Social scoring has become prevalent in some countries as a government tool for public administration, financial services, and employment decisions. A person’s social credit score might be based on their payment history, compliance with regulations, online behaviour, relationships with other low-scoring individuals, traffic violations, or even political opinions. If the score falls below a threshold, the person faces concrete restrictions: they cannot obtain credit or loans, cannot enrol their children in school, cannot book travel on flights or trains, or cannot obtain employment in certain sectors. The restrictions cascade, making it impossible for the person to build a normal life.
The EU considers this practice unacceptable regardless of who conducts it or their stated intentions. The prohibition covers social scoring by government agencies and by private companies equally. A private company running a social scoring system to assess customer creditworthiness based on character is just as prohibited as a government running one. The rationale is that social scoring systems lack transparency about how scores are calculated, deny people meaningful opportunity to contest their scores or understand how to improve them, and concentrate power in the hands of whoever controls the algorithm.
There is an important distinction between social scoring and other rating systems that organisations can lawfully use. A credit score predicting credit risk based on payment history and financial conduct is not prohibited. A job applicant evaluation system measuring job-relevant competencies and work experience is not prohibited. A reputation system on an e-commerce platform showing seller transaction history and customer feedback is not prohibited. A healthcare system assessing patient medical risk is not prohibited. The distinction is that prohibited social scoring reduces a person’s score based on their social behaviour, character or personal values, not on their objective capability or conduct relevant to the service being restricted.
The harm social scoring causes is severe and lasting. A low social credit score can trap a person in cycles of disadvantage, making it harder to access credit, employment and services no matter what they do to improve themselves. The score becomes self-perpetuating: it predicts exclusion, which causes further harm that validates the initial score. The person may not know why they are scored low, may not be able to challenge the score through any transparent process, and may not be able to understand what behaviour would improve it.
The prohibition applies to any AI system that produces this effect, including hybrid systems that combine AI scoring with human judgment afterwards. If the AI system produces a score and that score is used to cause the restriction, even if humans review it later, the core practice is prohibited. The timing and nature of human review does not change the analysis.
There is no exception to this prohibition for any organisation, public or private. There is no exemption for law enforcement. There is no exception based on accuracy of the score or fairness of the process. The complete prohibition on social scoring by any body reflects the EU’s considered judgment that this form of AI use is fundamentally incompatible with European values of human dignity and equality before public administration.
Examples of prohibited social scoring include: a government system that scores citizens based on legal compliance, political engagement, social media activity and perceived civic contribution, then restricts access to travel, credit and public services based on scores; a private financial services company that creates a “trustworthiness score” based on payment history, online behaviour, social media activity and relationships with other low-scoring individuals, then uses that score to restrict credit access; a public employment agency that scores job applicants based on perceived reliability inferred from background data, social connections and demographic profile, then prioritises certain applicants for training. In contrast, a system that assesses credit risk based on income and payment history, that evaluates job candidate capability based on skills and experience, or that screens loan applicants based on objective financial factors would not be prohibited social scoring.