At a glance

  • Since February 2, 2025, the EU AI Act has banned “real-time” remote biometric identification in publicly accessible spaces for law enforcement, except where strictly necessary for one of three objectives, under the conditions in Article 5(2) to (7), and only where national law provides for it.
  • From December 2, 2027, the Annex III law enforcement uses are high-risk; most public-authority systems already on the market must comply by August 2, 2030. Article 26(10) adds an authorisation rule for post-remote biometric identification.
  • In the US, of the seven state laws this guide verified (not a complete list), Washington, Utah, Montana, Maryland and Massachusetts allow police facial recognition under conditions; Vermont bars it, with two exceptions; and in Virginia, the 2022 law that authorized police use expired on July 1, 2026.

What rules limit police use of AI?

In the EU, the AI Act bans police use of real-time remote biometric identification, such as facial recognition, in publicly accessible spaces except where strictly necessary for one of three objectives, subject to binding authorisation from a judicial or independent administrative authority, and only where national law provides for it. It also bans crime-risk predictions about individuals based solely on profiling or personality traits, and treats listed police uses as high-risk from December 2, 2027. In the US, where this guide found no federal statute that specifically governs police use of AI or facial recognition as of November 16, 2026, the limits include state facial recognition statutes, local ordinances, federal policy binding only federal agencies, and the courts’ evidence rules.

Which rules apply, and when?

Rules on police use of AI, as of November 16, 2026
RuleJurisdictionWhat it limitsBinding?Status or date
EU AI Act Article 5(1)(h) and 5(2) to (7)EUUse of real-time remote biometric identification in publicly accessible spaces for law enforcement: three objectives only; prior judicial or independent administrative authorisation, or, in a duly justified situation of urgency, a request without undue delay and at the latest within 24 hours; impact assessment and registration; notification; annual reportsBinding; usable only where national law provides for itSince February 2, 2025
EU AI Act Article 5(1)(d) and (e)EUCrime-risk assessments of individuals based solely on profiling or personality traits; untargeted scraping of facial images to build or expand facial recognition databasesBindingSince February 2, 2025
EU AI Act Annex III points 1 and 6, Articles 26(10), 27 and 49(4)EUHigh-risk duties for listed police uses; authorisation for post-remote biometric identification; fundamental rights impact assessment by public bodiesBindingDecember 2, 2027; August 2, 2030 for public-authority systems already on the market; December 31, 2030 for components of the Annex X large-scale IT systems placed on the market or put into service before August 2, 2027 (Article 111(1))
Law Enforcement Directive (EU) 2016/680, Articles 10 and 11EUBiometric processing for unique identification only where strictly necessary and on a listed ground; decisions based solely on automated processing that produce an adverse legal effect or significantly affect the person, unless authorised by lawBinding through national lawIn force
OMB Memorandum M-25-21US federal agenciesMinimum practices, including an impact assessment, before deploying high-impact AI; many law enforcement uses are presumed high-impact where the AI is a principal basis for a decision or actionBinding on agencies only; does not cover AI used as a component of a National Security System; elements of the Intelligence Community are not required to apply its minimum practices, and pilot programs that meet its conditions are exempt from them; a Chief AI Officer may waive one or more minimum practices for a specific covered AI application after a written determinationApril 3, 2025
RCW 43.386WashingtonState and local agency facial recognition: accountability report; no ongoing surveillance, real-time or near real-time identification, or persistent tracking without a warrant, exigent circumstances, or a court order for missing or deceased persons; no sole basis for probable causeBindingEffective July 1, 2021
Utah Code 77-23eUtahGovernment facial recognition on image databases: law enforcement requests only, for listed purposes, with a second trained reviewerBindingEffective May 5, 2021
MCA 44-15-101 et seq.MontanaBans continuous facial surveillance; searches for listed purposes with a warrant, or a court order for missing or deceased persons, and a warrant within 24 hours after an emergency search; no sole basis for probable causeBindingCh. 781, Laws of 2023; definitions in 44-15-103 amended in 2025 (Chs. 226 and 484)
Criminal Procedure 2-501 to 2-510MarylandListed crimes only; no live or real-time identification; trained verification; results may be introduced in a criminal proceeding only to establish probable cause or positive identification for a warrant or at a preliminary hearing, never as the sole basis, and only where additional, independently obtained evidence supports them; not in juvenile delinquency proceedings; results obtained in violation, and derived evidence, generally excluded (2-503(c)); annual auditBindingEffective October 1, 2024
M.G.L. c. 6, s. 220MassachusettsSearches only by written request to the Registry of Motor Vehicles, State Police or FBI, to execute a court order, or without one to identify a deceased person or in an emergencyBindingEffective July 1, 2021 (St. 2020, c. 253, sections 26 and 122); text read November 16, 2026
2020 Acts and Resolves No. 166, Sec. 14; 2021 Act No. 17VermontMoratorium on law enforcement use of facial recognition, except uses permitted for drones under 20 V.S.A. 4622 and, on legally seized electronic media, child sexual exploitation investigationsBindingAct 166 signed October 7, 2020 (Sec. 33 sets October 1, 2020); Act 17 effective on passage, signed May 4, 2021
Va. Code 15.2-1723.2 and 23.1-815.1VirginiaLocal and campus police may not purchase or deploy facial recognition unless a statute expressly authorizes it; a statute that does not refer to facial recognition does not count; the State Police section, 52-4.5, expired July 1, 2026BindingVersion effective July 1, 2026

When may EU police use real-time facial recognition?

In publicly accessible spaces, only where strictly necessary for one of three objectives: the targeted search for specific victims of abduction, trafficking in human beings, or sexual exploitation, and the search for missing persons; the prevention of a specific, substantial, and imminent threat to life or physical safety, or a genuine and present or genuine and foreseeable threat of a terrorist attack; or the localisation or identification of a person suspected of having committed a criminal offence, for the purpose of conducting a criminal investigation or prosecution or executing a criminal penalty, for offences listed in Annex II and punishable in the member state concerned by a custodial sentence or detention order for a maximum period of at least four years (Article 5(1)(h)).

Each use may serve only to confirm the identity of the specifically targeted individual. The authority must have completed a fundamental rights impact assessment under Article 27 and registered the system under Article 49, although in duly justified cases of urgency use may begin before registration if registration follows without undue delay (Article 5(2)). Each use needs prior authorisation from a judicial authority, or an independent administrative authority whose decision is binding; in a duly justified situation of urgency, authorisation must be requested without undue delay and at the latest within 24 hours, and if it is rejected the use must stop immediately and all data, results, and outputs must be discarded and deleted (Article 5(3)). No decision that produces an adverse legal effect on a person may be taken based solely on the system’s output (Article 5(3)).

Each use must be notified to the relevant market surveillance authority and the national data protection authority, without sensitive operational data (Article 5(4)), and those authorities report annually to the Commission (Article 5(6)). The route exists only where a member state’s national law provides for it, and member states may adopt more restrictive laws (Article 5(5)).

Each member state decides to what extent its public authorities can be fined (Article 99(8)). The EU AI Act penalties and prohibited practices guide covers the fine tiers.

Which other EU AI Act bans reach police?

  • Predictive policing of individuals. Article 5(1)(d) bans risk assessments of individuals to assess or predict the risk of committing a criminal offence, based solely on profiling or on assessing personality traits and characteristics. It does not cover AI systems used to support the human assessment of a person’s involvement in a criminal activity, which is already based on objective and verifiable facts directly linked to a criminal activity.
  • Scraped face databases. Article 5(1)(e) bans creating or expanding facial recognition databases through the untargeted scraping of facial images from the internet or CCTV footage.
  • Categorisation and emotion recognition. The Article 5(1)(g) biometric categorisation ban does not cover labelling or filtering lawfully acquired biometric datasets, including in the area of law enforcement (guidelines, paragraph 284), and the Article 5(1)(f) emotion recognition ban covers only the workplace and education institutions. Both uses can be high-risk under Annex III point 1.

Which police AI uses are high-risk in the EU, and what changes in December 2027?

Annex III point 6 lists AI used by or on behalf of law enforcement authorities, or Union bodies supporting them, for: (a) assessing the risk of a person becoming a victim; (b) polygraphs and similar tools; (c) evaluating the reliability of evidence; (d) assessing the risk of offending or re-offending not solely on the basis of profiling, or assessing personality traits or past criminal behaviour; and (e) profiling in detection, investigation, or prosecution. These uses are listed “in so far as their use is permitted under relevant Union or national law,” so a listing does not make a use lawful (recital 63). The high-risk classification guide covers the full annex.

The high-risk rules apply from December 2, 2027 for stand-alone Annex III systems. Apart from components of the Annex X large-scale IT systems (Article 111(1)), a system placed on the market or put into service before then is covered only if its design changes significantly from that date, but providers and deployers of high-risk systems intended to be used by public authorities must comply by August 2, 2030 (Article 111(2)). Two provisions reach police:

  • Post-remote biometric identification (Article 26(10)). In an investigation for the targeted search of a person suspected or convicted of having committed a criminal offence, the deployer must request authorisation, ex ante or without undue delay and no later than 48 hours, from a judicial authority or an administrative authority whose decision is binding and subject to judicial review, except for the initial identification of a potential suspect based on objective and verifiable facts directly linked to the offence. Each use must be limited to what is strictly necessary for the investigation of a specific criminal offence, and if authorisation is rejected the use must stop and the linked personal data must be deleted. The system may never be used in an untargeted way, without any link to a criminal offence, a criminal proceeding, a genuine and present or genuine and foreseeable threat of a criminal offence, or the search for a specific missing person. No decision with an adverse legal effect may rest solely on its output, and each use must be documented in the relevant police file.
  • Impact assessment and registration. Before deploying a high-risk system referred to in Article 6(2), other than a critical infrastructure system under Annex III point 2, deployers that are bodies governed by public law or private entities providing public services must assess the impact on fundamental rights that the use of the system may produce (Article 27(1)). Systems under points 1, 6, and 7 in the areas of law enforcement, migration, asylum, and border control management register in a secure non-public section of the EU database (Article 49(4)).

Does the GDPR apply to police AI?

Not to criminal law enforcement processing. The GDPR does not apply to processing by competent authorities for the prevention, investigation, detection or prosecution of criminal offences or the execution of criminal penalties, including the safeguarding against and the prevention of threats to public security (Article 2(2)(d)). The Law Enforcement Directive, applied through national law, covers it instead, and its Articles 10 and 11 limit biometric processing and decisions based solely on automated processing. The GDPR AI compliance guide covers the GDPR side.

What does US federal law require?

For federal agencies, OMB Memorandum M-25-21, issued April 3, 2025, requires an AI impact assessment before deploying any high-impact AI use case. Its Section 6 lists purposes for which AI that serves as a principal basis for an agency decision or action is presumed to be high-impact. In law enforcement contexts they include risk assessments about individuals, identification of criminal suspects, forecasts of crime, biometric identification such as iris, facial, fingerprint, or gait matching, and determinations related to recidivism, sentencing, parole, or bail; a separate category covers one-to-many biometric identification in publicly accessible spaces. The list is not exhaustive. The memo does not cover AI used as a component of a National Security System, elements of the Intelligence Community are not required to apply its minimum practices, and pilot programs that meet its conditions are exempt from them. It binds agencies, not vendors or state and local police.

A December 2024 Biometric Technology Report by DHS, DOJ and the White House Office of Science and Technology Policy, now archived, says the Deputy Attorney General announced an interim policy in December 2023. As the report describes it, the policy bars First Amendment activity as the sole basis for use and states that results alone may not be relied upon as the sole proof of identity. The report was prepared under Executive Order 14074, which Executive Order 14148 revoked on January 20, 2025. This guide could not confirm whether the interim policy remains in effect.

Which states restrict police facial recognition?

This guide verified seven state laws from their text; the list is not complete.

  • Washington (RCW 43.386). An agency must publish an accountability report, after public comment and community consultation, at least 90 days before operational use. It may not use facial recognition for ongoing surveillance, real-time or near real-time identification, or persistent tracking unless a warrant is obtained, exigent circumstances exist, or a court order is obtained for missing or deceased persons, and police may not use the results as the sole basis to establish probable cause. Use on a criminal defendant must be disclosed to that defendant in a timely manner before trial.
  • Utah (77-23e). Only a law enforcement agency may request a comparison, for investigating a felony, a violent crime, or a threat to human life, or identifying a deceased, incapacitated or at-risk person. A trained employee must get a second opinion before returning a probable match, and police must tell the prosecutor in writing whether facial recognition was used.
  • Montana (44-15). No state or local agency may use continuous facial surveillance. Police may search to investigate a serious crime where there is probable cause that an unidentified person in an image committed it, was a victim, or witnessed it, or to locate or identify a missing, endangered or deceased person, with a warrant, or a court order for missing or deceased persons; after a search in an emergency posing an imminent threat to a person, they must obtain a warrant within 24 hours. Results may not be the sole basis for probable cause, and data obtained in violation of the chapter is inadmissible.
  • Maryland (Criminal Procedure 2-501 to 2-510). In a criminal investigation, use is limited to listed crimes, live or real-time identification is barred, and a result must be independently verified by a trained individual before any use (2-503). Results may not be introduced in a juvenile delinquency proceeding, and may be introduced in a criminal proceeding only to establish probable cause or positive identification for a warrant or at a preliminary hearing, never as the sole basis, and only where additional, independently obtained evidence supports them (2-502). The State must disclose use in discovery (2-504), and each agency must complete an annual compliance audit (2-505).
  • Massachusetts (c. 6, s. 220). Police may search only through a written request to the Registry of Motor Vehicles, the State Police or the FBI, to execute a court order based on specific and articulable facts, or without an order to identify a deceased person or in an emergency involving substantial risk of harm.
  • Vermont. Until the General Assembly authorizes it, a law enforcement officer may not use facial recognition, or information acquired through it, unless the use would be permitted for drones under 20 V.S.A. 4622 (2020 Act 166, Sec. 14). Act 17, signed May 4, 2021, allows it in investigations into sexual exploitation of children, confined to legally seized electronic media.
  • Virginia. The 2022 act that authorized local, campus and State Police use provided that it would expire on July 1, 2026 (2022 Acts, c. 737). As the Code of Virginia read on November 16, 2026, local and campus police may not purchase or deploy facial recognition unless a statute expressly authorizes it, and a statute that does not refer to facial recognition does not count. This guide found no 2026 act that replaced the expired authority.

San Francisco’s Ordinance 107-19 (June 2019) made it unlawful for any city department to obtain, retain, access, or use face recognition technology or information obtained from it (Administrative Code 19B.2(d)). Ordinance 286-19 (December 2019) limited the ban to technology on City-issued software, products, or devices, and information obtained from it, and allowed departments to possess unused, manufacturer-installed features under conditions. The current 19B.2(d), amended through 2025, keeps that limit, subject to 19B.2(e) to (l).

Does Illinois BIPA reach police?

Not directly. BIPA regulates private entities, and State and local government agencies are not private entities. Section 25 also excludes contractors, subcontractors, and agents of a State agency or local unit of government when working for it. The Illinois AI laws guide covers BIPA’s duties.

Can police rely on AI-generated or AI-enhanced evidence in court?

In federal court, Rule 702 and authentication under Rule 901 still govern AI-generated exhibits offered today. Proposed Federal Rule of Evidence 707, which has not been adopted, would apply the Rule 702 reliability standards to machine-generated evidence offered without an expert witness. On May 7, 2026, the Advisory Committee on Evidence Rules did not recommend action, revised the draft, and decided not to republish it yet. Its October 15, 2026 agenda book lists artificial intelligence and proposed Rule 707; this guide found no published record of the outcome.

What should agencies and vendors do?

  • Agencies: before a facial recognition search, confirm the statute lists the purpose and obtain any warrant or court order it requires
  • Agencies: have a trained person verify each match, and never use a match as the sole basis for probable cause
  • Agencies in the EU: confirm national law provides for any real-time use, and plan authorisation and reporting for real-time and post-remote identification
  • Vendors: meet state vendor duties, such as Washington’s independent testing and bias disclosure and Montana’s written privacy policy and written consent

The AI impact assessment guide compares the Article 27 assessment with the one OMB M-25-21 requires.

This guide is general information as of November 16, 2026, not legal advice. Confirm each point against the statute or regulation in force and with counsel or the regulator.

Common questions

Can police use real-time facial recognition in the EU?

In publicly accessible spaces, only where strictly necessary for one of three objectives: the targeted search for specific victims of abduction, trafficking in human beings, or sexual exploitation, and the search for missing persons; the prevention of a specific, substantial, and imminent threat to life or physical safety, or a genuine and present or genuine and foreseeable threat of a terrorist attack; or the localisation or identification of a person suspected of having committed a criminal offence, for conducting a criminal investigation or prosecution or executing a criminal penalty, for offences listed in Annex II and punishable in the member state concerned by a custodial sentence or detention order for a maximum period of at least four years. Each use needs prior authorisation from a judicial authority or an independent administrative authority whose decision is binding, which in a duly justified situation of urgency must be requested without undue delay and at the latest within 24 hours; if authorisation is rejected, the use must stop immediately and all data, results, and outputs must be discarded and deleted. Each use may serve only to confirm the identity of the specifically targeted individual, the authority must have completed a fundamental rights impact assessment and registered the system (in duly justified cases of urgency, registration may follow without undue delay), and the route exists only where a member state's national law provides for it (Article 5(1)(h), (2), (3) and (5)).

Is there a US federal law on police use of facial recognition?

This guide found no federal statute that specifically governs police use of AI or facial recognition as of November 16, 2026. OMB Memorandum M-25-21 lists law enforcement uses, including biometric identification and identification of criminal suspects, among the purposes presumed to be high-impact where the AI serves as a principal basis for an agency decision or action, but it binds federal agencies, not vendors or state and local police.

Can Virginia police use facial recognition?

The 2022 act that authorized local, campus and State Police use expired on July 1, 2026 (2022 Acts, c. 737). As the Code of Virginia read on November 16, 2026, local and campus police may not purchase or deploy facial recognition unless a statute expressly authorizes it, and a statute that does not refer to facial recognition does not count. That rule does not cover the State Police, whose own section, 52-4.5, expired with the 2022 act. This guide found no 2026 act that replaced the expired authority.

By industry: all 8 industry pages

Sources

  1. Regulation (EU) 2024/1689 (Artificial Intelligence Act), Official Journal of the European Union, July 12, 2024.
  2. Regulation (EU) 2024/1689, consolidated text as of July 27, 2026, as amended by Regulation (EU) 2026/1744, EUR-Lex, Articles 3(41) to (46), 5, 26(10), 27, 49(4), 99, 111 and 113 and Annexes II, III and X, accessed November 16, 2026.
  3. Regulation (EU) 2026/1744 (Digital Omnibus on AI), Official Journal of the European Union, July 24, 2026.
  4. Directive (EU) 2016/680 on the protection of natural persons with regard to the processing of personal data by competent authorities for the purposes of the prevention, investigation, detection or prosecution of criminal offences or the execution of criminal penalties (Law Enforcement Directive), Official Journal of the European Union, May 4, 2016.
  5. Regulation (EU) 2016/679 (General Data Protection Regulation), Official Journal of the European Union, May 4, 2016, Article 2(2)(d).
  6. Communication from the Commission, Commission Guidelines on prohibited artificial intelligence practices established by Regulation (EU) 2024/1689 (AI Act), C(2025) 5052 final, European Commission, July 29, 2025.
  7. OMB Memorandum M-25-21, Accelerating Federal Use of AI through Innovation, Governance, and Public Trust, Office of Management and Budget, April 3, 2025, Sections 4 and 6.
  8. Biometric Technology Report, submitted in fulfillment of Section 13(e) of Executive Order 14074 (PDF, archived), U.S. Department of Homeland Security, U.S. Department of Justice, and White House Office of Science and Technology Policy, December 2024, section 4.2.2.
  9. Executive Order 14148, Initial Rescissions of Harmful Executive Orders and Actions, 90 FR 8237 (PDF), Federal Register, January 28, 2025, section 2(vv).
  10. Chapter 43.386 RCW, Facial recognition, Washington State Legislature, accessed November 16, 2026.
  11. Utah Code Title 77, Chapter 23e, Government Use of Facial Recognition Technology, Utah State Legislature, accessed November 16, 2026.
  12. Montana Code Annotated 2025, Title 44, Chapter 15, Part 1, Facial Recognition for Government Use, Montana Legislature, accessed November 16, 2026.
  13. Maryland Code, Criminal Procedure 2-503 (with 2-502, 2-504, 2-505 and 2-510), Maryland General Assembly, accessed November 16, 2026.
  14. SB 182 (2024), Criminal Procedure, Facial Recognition Technology, Chapter 808, Maryland General Assembly, effective October 1, 2024.
  15. General Laws, Part I, Title II, Chapter 6, Section 220, Massachusetts Legislature, accessed November 16, 2026.
  16. Acts of 2020, Chapter 253, An Act Relative to Justice, Equity and Accountability in Law Enforcement in the Commonwealth, sections 26 and 122, Massachusetts Legislature, approved December 31, 2020.
  17. 2020 Acts and Resolves No. 166 (S.124), Sec. 14 (PDF), Vermont General Assembly, signed October 7, 2020.
  18. Act No. 17 (H.195), Use of facial recognition technology by law enforcement in cases involving sexual exploitation of children (PDF), Vermont General Assembly, signed May 4, 2021.
  19. Code of Virginia 15.2-1723.2, Facial recognition technology; approval, Virginia Legislative Information System, accessed November 16, 2026.
  20. Code of Virginia 23.1-815.1, Facial recognition technology; approval, Virginia Legislative Information System, accessed November 16, 2026.
  21. Code of Virginia 52-4.5 (effective until July 1, 2026), Virginia Legislative Information System, accessed November 16, 2026.
  22. 2022 Acts of Assembly, Chapter 737, Virginia Legislative Information System, accessed November 16, 2026.
  23. 2026 Legislative Update, Regular Session Winter/Spring 2026, Master List (PDF), Virginia Commonwealth’s Attorneys’ Services Council, published by the Virginia Department of Criminal Justice Services, 2026.
  24. Ordinance No. 107-19, Administrative Code, Acquisition of Surveillance Technology (PDF), San Francisco Board of Supervisors, finally passed June 4, 2019.
  25. Ordinance No. 286-19, Administrative Code, Acquisition, Retention, and Use of Surveillance Technology (PDF), San Francisco Board of Supervisors, finally passed December 17, 2019, Section 19B.2(d) to (h).
  26. Ordinance No. 50-26, Police Surveillance Technology Policy for Electronic Location Tracking Devices (PDF), San Francisco Board of Supervisors, finally passed April 7, 2026, Section 1(b) (Proposition E).
  27. San Francisco Administrative Code, Section 19B.2, American Legal Publishing, amended through Ordinance 284-24 (effective January 19, 2025), read November 17, 2026, subsections (d) to (l).
  28. Biometric Information Privacy Act (740 ILCS 14), Illinois General Assembly, accessed November 12, 2026.
  29. Report of the Advisory Committee on Evidence Rules (PDF), Judicial Conference of the United States, May 17, 2026.
  30. Advisory Committee on Evidence Rules, Agenda Book, October 15, 2026 (PDF), United States Courts.