Non-discrimination and equality
The application of AI system in law enforcement and public security also raises concerns about algorithmic bias leading to discrimination. For example, it has been shown in several cases that bias is embedded in facial recognition systems, resulting in the misidentification of suspects and, in some instances, the wrongful incarceration of innocent individuals.[1] Moreover, AI tools to enhance facial recognition technology, when trained on inadequate or skewed datasets, may reduce the accuracy of identification for some groups.[2] States should exercise caution with respect to identifying, assessing, preventing, and mitigating risks of discrimination arising from the use of, for example, facial recognition technologies or remote biometric identification systems in the law enforcement and security sectors. States may assess whether new regulations are necessary or if specific measures, including explicit prohibitions, should be implemented to prevent discrimination.[3]
In the context of prison and probation services, Recommendation CM/Rec(2024)5 underlines that safeguards must be in place to prevent discrimination, ensure procedural fairness, and uphold human dignity, ensuring that AI-driven prison management remains compatible with fundamental rights and the rule of law. When developing AI and related digital technologies in order to increase the accuracy and objectivity of risk assessment, the challenges of algorithmic biases and quality and representativeness of data should be addressed. Sensitivity to all kinds of diversity, including to gender perspective, different forms of disabilities and multiculturalism, should inform the design and use of risk assessment tools in order to avoid any discrimination. When such tools are used for the personalisation of treatment and reintegration plans, this should be done with care to avoid biases. The use of such tools should not replace regular face‑to‑face human contact between professionals and the offenders, including, where necessary, the work with their families and children, in line with Recommendation CM/Rec(2018)5 of the Committee of Ministers to member States concerning children with imprisoned parents.
[1] GEC/CDADI Study (2023), pp. 22-23. More examples can be found in Resolution 2342 (2020) “Justice by algorithm – The role of artificial intelligence in policing and criminal justice systems’, paragraph 7.
