The development and use of AI systems in law enforcement and public security could present particular human rights risks. This is because of the strong human rights impact of actions or decisions that might be taken based on AI systems output such as surveillance, search and seizure, or arrest and detention. The use of AI systems in this sector may interfere, in particular with Articles 5 (Right to liberty and security), 8 (Right to respect for private and family life), 10 (Freedom of expression), and 11 (Freedom of assembly and association) of the ECHR, as well Article 2 of Protocol no. 4 to the ECHR (Freedom of movement).[1] States may justify interference with Articles 8, 10 and 11 ECHR by the legitimate aims listed in the texts of these articles which include national security, public safety, or the prevention of disorder or crime.

 

The right to liberty and security

Predictive policing systems make estimations and predictions that may be turned into concrete actions or decisions by the criminal justice system, including on arrest and detention. Due to the decisions that could be made based on such systems output, Article 5 ECHR (the right to liberty and security) issues may arise. Decisions on arrest or detention must be based on reasonable suspicion based on verifiable and objective facts directly linked to a criminal activity.[2]  Should information provided by predictive policing systems be used to corroborate reasonable suspicion for a decision or arrest and detention, explainability and interpretability issues (the “black box problem”) concerning AI systems may pose difficulties to meet the criteria required for verifiability and objectivity. Predictive policing methods must not lead to unlawful decisions on deprivation of liberty. Such operations carried out by public authorities must be lawful, necessary, and proportionate to their intended purposes and be based on clear, foreseeable, and accessible domestic law, pursuing a legitimate aim while ensuring adequate safeguards.[3]

 

[1] Applicable to those States that have ratified this Protocol.

[2] Akgün v. Turkey, No. 19699/18, 20 July 2021, §§ 156 and 175.

[3] See also Article 5.(d) of the EU AI Act, which concerns the use of AI systems in policing.