Facial recognition technology (FRT) is increasingly used by law enforcement to identify and arrest suspects, yet regulations across political contexts are inconsistent and lack clarity. This comparative case study assesses FRT regulation in the criminal justice systems of five developed democracies (US, Canada, Germany, Italy, France), reporting substantial variation.
Research Objectives
- Identify the ethical challenges of FRT in arrests and whether regulation mitigates them.
- Document existing FRT regulations in each of the five countries.
- Compare how those regulations vary across contexts.
- Propose principles for an ethical framework for developing, deploying, and regulating FRT.
Methodology
- Comparative case study of five developed democracies, purposively selected for varied approaches to technology governance; authoritarian regimes were excluded.
- Regulations were compared on legal compliance, ethical alignment (privacy, accountability, consent), and effectiveness in implementation.
- Each case draws on secondary sources (statutes, regulator guidance, scholarship, news reporting) and includes examples of enforcement use and public reactions.
Key Findings
- Political structure tracks fragmentation: the federal US has no national FRT regulation, only varied state rules; Canada has a similar patchwork but, unlike the US, a national privacy commission that issued FRT guidance; Germany, federal with a stronger central government, is less fragmented.
- Unitary Italy and France have uniform national policy but diverge: France expanded FRT use significantly during the 2024 Paris Olympics; Italy's moratorium runs through 2025, excepting judicial authorities and fighting crime.
- Consent rules are largely similar: not required in criminal justice, typically required elsewhere. Canada permits use without consent where need is explicit (passport control) but requires it for cross-matching.
- NIST testing in late 2019 found most of nearly 200 algorithms had accuracy varying by demographic group: Asians, African Americans, American Indians, women, children, and the elderly were misidentified more often than white individuals, men, and middle-aged adults respectively. Some algorithms lacked detectable bias, so the risk can be mitigated during development.
- Six documented wrongful arrests of Black individuals in the US are connected to FRT, and comprehensive data on the extent of use and the total number of such arrests is not available. Germany's 2017 Berlin Südkreuz prototype, on consenting volunteers, produced accuracy the authors call underwhelming.
Recommendations
- Embed privacy by design and by default in national FRT regulation, with DPIAs for high-risk uses.
- The authors note significant risks and concerns if law enforcement uses FRT output as the sole basis for an arrest, given the potential for inaccuracy or bias.
- Evaluate consent mechanisms for efficacy in protecting autonomy and rights, not just presence.
- Improve data collection on law enforcement FRT use and wrongful arrests.
- Develop an adaptable FRT-specific ethical framework for criminal justice, built on primary data from policymakers, legal experts, and the public.
Key Takeaways
- Regulatory responses differ substantially even among established democracies, and the authors find no AI ethics framework specifically addressing FRT in criminal justice.
- Federal structures allow policy experimentation, but inconsistent rules on privacy, accountability, and consent, combined with the technology's equity pitfalls, can produce significant inequalities in policing. Unitary systems carry the opposite risk: faster uniform adoption nationwide.
- The study covers only five developed democracies and calls for wider comparison. For Thailand and ASEAN it raises questions: whether biometric oversight has a clear locus, and whether consent rules separate criminal justice from other public use.
References
Robles, P., Mallinson, D. J., Best, E., Devaney, C., & Azevedo, L. (2025). Global perspectives on regulating facial recognition technology utilization for criminal justice arrests. Global Public Policy and Governance, 5, 186–204. https://doi.org/10.1007/s43508-025-00117-9
Full text (Open Access): https://link.springer.com/article/10.1007/s43508-025-00117-9
