By Michele Evans / NYweeklyRecord.com / Date: 11/20/2026
Category: NYPD Accountability / Surveillance / Civil Rights / Government Oversight
NEW YORK CITY, NY –
New York City’s police watchdog found that the NYPD ran about 26,000 facial-recognition analyses in 2025 while key parts of the system remained inconsistent, poorly documented or out of step with the department’s own rules.
The Department of Investigation’s Office of the Inspector General for the NYPD said the department largely complied with the city’s Public Oversight of Surveillance Technology Act. That bottom-line finding did not erase a long list of failures. The inspector general issued 17 recommendations aimed at making searches more transparent, consistent and auditable.
The scale matters. Logs reviewed by the inspector general showed that the NYPD’s Facial Identification Section conducted approximately 26,000 analyses in a single year. A possible match is not proof that a person committed a crime. Department policy requires investigators to treat it as a lead that must be checked against other evidence.
DOI found that two branches of the NYPD did not always use the same system to record searches. The Facial Identification Section logged work through the department’s Enterprise Case Management System. Intelligence Division personnel used email and separate paper or digital logs through 2025, contrary to the Patrol Guide.
That split created a risk that different units could search the same case without knowing what the other unit had done. It also made it harder to determine whether searches, results and reviews were preserved accurately. In one Intelligence Division log, 33 percent of entries listed the same investigator as both the requestor and the person who conducted the search, even though facial-recognition investigators did not carry their own cases.
Supervision was another weak point. DOI found that roughly one-third of Intelligence Division searches recorded in 2024 and 2025 were marked “no” for supervisor review. The NYPD told the inspector general that its rules require supervisory review of any possible match. The report also found that units did not document peer-review outcomes in a way that allowed reliable auditing.
Outside searches raised a separate accountability problem. NYPD records showed 98 approved requests to use U.S. Customs and Border Protection facial-recognition systems between October 2023 and October 2025. Every request in the logs came from the Intelligence and Counterterrorism Bureau and was approved by an assistant commissioner.
DOI found that the approval route did not match the facial-recognition policy in effect at the time. The records did not explain why each request was granted and did not show whether any requests were denied. The inspector general recommended that the NYPD identify who may approve outside searches and record the decision and reasoning in every case.
The report grew out of a 2025 Legal Aid Society referral alleging that NYPD personnel had gone outside approved photo repositories and relied on searches conducted by other agencies. Two criminal cases cited by Legal Aid were dismissed and sealed, limiting DOI’s access to the underlying records. The inspector general said it could not make findings about those individual cases.
One court ruling, however, documented a policy violation. In People v. Zuhdi A., a judge found that an FDNY facial-recognition search that crawled the open web went beyond what NYPD policy permits. DOI cited that ruling while recommending formal rules for searches performed by outside entities and for officers’ use of those results.
The audit also found that the department’s public policy did not name all three facial-recognition products in use, inaccurately described the photo repositories attached to those products and lacked clear guidance on when investigators should choose one system over another. DOI called for standardized logs, written rules for modifying probe images, mandatory blind peer review, recurring training and annual audits.
The NYPD had rejected two related inspector-general recommendations in 2022. The City Council later wrote those requirements into law in 2025. The department told DOI that it intends to conduct the required facial-recognition audit each calendar year beginning in 2026.
The report notes that the NYPD made some policy changes after receiving a draft and cooperated with the review. It does not say that the department accepted every recommendation. That leaves the next accountability test in City Hall and at police headquarters: which safeguards will be adopted, when they will take effect and whether the public will be able to verify compliance.
Privacy advocates want lawmakers to go further. The Surveillance Technology Oversight Project cited the 26,000 searches and the CBP requests in renewing its call to ban police use of facial recognition. That is an advocacy position, not a finding in the DOI report.
For everyday New Yorkers, the issue is not limited to software. A search can influence who is questioned, placed in a lineup or arrested. When the records are inconsistent and outside searches are poorly explained, a person caught by a bad lead may have no simple way to reconstruct how the government reached them.
Facial recognition may produce investigative leads. Oversight must show how those leads were created, reviewed and tested. The inspector general’s report shows that New York City still cannot answer every one of those questions from its own records.
Sources
NYC Department of Investigation: An Assessment of NYPD’s Compliance with the POST Act


Leave a Reply