The Troy Police Department produced a case-by-case tracking log to argue against the City Council’s proposed 48-hour data deletion window for Flock cameras. Every closed arrest in that document came from a search conducted within 48 hours of the plate being read.
The log lists 30 cases with case numbers, investigation types, and outcome notes. Eleven of those 30 cases, 37 percent, required data older than 48 hours. Not one produced a closed arrest. The cases requiring older data are open investigations, cases where victims declined to continue, searches that could not identify a suspect, and searches run on behalf of other agencies. The Administration’s argument for longer retention rests entirely on cases that did not result in arrests.
That document was provided directly to Council Member Noreen McKee. It was not included in the FOIL response filed July 27, which asked for all reports prepared for or delivered to the Mayor or City Council.
The broader audit, a one-page narrative and two spreadsheets covering May 23 through July 23, 2026, was produced under the May 19 agreement between the Mantello Administration and the City Council. That agreement required annual audits and established a 60-day collaborative period during which the police department would provide data to the Council. The window expired July 18. On July 24, Council President Sue Steele told WAMC she had not yet seen the results. The Council and the FOIL requestor both received the audit on July 29, eleven days after the deadline.
What the narrative says
The TPD’s one-page summary describes the cameras as a “controlled, high-yield asset” that provides “exceptional return on investment.” It reports 826 searches executed across 81 unique queries over the 60-day period, with 42.6% of authorized personnel conducting searches. Of 30 tracked cases: 7 arrests, 2 warrants, 9 active investigations, and 12 cases categorized as “operational resolution,” a category that includes cleared unfounded reports, inter-agency intelligence sharing, and cases where victims declined prosecution. The narrative does not break down how many cases fall into each sub-category.
The narrative’s final paragraph argues against the Council’s proposed 48-hour data deletion window. It cites a July 4 hit-and-run investigation that required data three to four weeks old, then cautions that the 60-day sample window is too short to draw conclusions about typical data retention needs. The document uses its showcase case to argue for longer data retention, then argues the same case is too exceptional to draw conclusions from.
What the spreadsheets show
The raw search logs tell a different story than the narrative summary. The two spreadsheets total 824 rows, not 826. Each row contains a Flock-generated ID, the agency name, a search timeframe, a reason code, and a search type. Three fields are absent from every row: the name or badge number of the officer who ran the search, the case or incident number the search was tied to, and the name of any person who reviewed or certified the data. The audit was produced by the department about the department and signed by no one.
Traffic infractions. Fifteen searches in the first period are categorized as “Traffic Infraction,” including one conducted on behalf of the State Police. The cameras were presented to the Council as tools for serious crime investigation. Traffic infractions are not serious crimes. The May 19 agreement did not address whether traffic enforcement is a permitted use. The category does not appear in the second-period log.
Inter-agency use is broader than disclosed. The narrative mentions “inter-agency intelligence sharing” as one element of the “operational resolution” category, without quantifying it. The logs show at least 25 searches run explicitly on behalf of other departments: Schenectady PD (2 searches), Watervliet PD (4), State Police (1), Cohoes (10), Colonie PD (7), and Albany (1). Troy is operating Flock as a regional search service. The scope was not disclosed in the narrative.
“No case.” Two searches in the first period list “Weapons Offense (Guns/Shots Fired)” as the reason, with the notation “No case.” Searches were conducted. No cases existed.
The search count is inflated. The narrative leads with 826 searches across 81 unique queries, an average of more than 10 searches per query. The July 4 hit-and-run the narrative uses as its central example is visible in the second-period log: at least seven overlapping time windows centered on July 5 between 1:00 and 1:30 AM account for between 40 and 50 searches. That single incident drives a significant share of the total. Without officer IDs or case numbers attached to each row, there is no way to verify how many distinct investigations the remaining searches represent.
Mobile searches. The search type field distinguishes standard lookups from mobile lookups, conducted from a phone or tablet in the field. In the first period, 151 of 312 searches (48%) were mobile. In the second period, 134 of 512 searches (26%) were mobile. The reason for the drop is not explained.
One entry without a case number
One entry in the tracking log carries no case number and no incident date. An individual described as “known to Police” was searched using 30 days of ALPR history based on an anonymous web tip. The reason code is “Narcotics/Firearm Possession.” No formal investigation was attached to the search. No arrest resulted. It is the entry with the longest data window and the least documentation.
What the audit cannot answer
The audit was designed to report activity, not to enable oversight. Without an officer name or badge number on each row, the department cannot use its own audit to detect whether any individual officer conducted an unusual number of searches. Without a case or incident number on each row, there is no way to verify that any given search was authorized by an open investigation. Without a named auditor, no one has certified the data.
The May 19 agreement gave the Council oversight authority over a $78,000-per-year contract. The audit produced to satisfy that agreement cannot answer the most basic accountability question: who ran this search, and what case authorized it?
What other jurisdictions produce
The fields missing from Troy’s audit are not technical limitations of the Flock system. Flock’s audit logs in other jurisdictions include officer name, plate searched, reason code, and the number of networks the query reached. TPD produced a version of the log without those fields. In July 2026, the Albany, Georgia Police Department arrested five officers after a Flock audit surfaced misuse: searches run for purposes unrelated to legitimate law enforcement investigations. That audit identified the officers by name. Troy’s audit cannot make the same determination because officer names are not in the produced records.
The Los Angeles Police Department’s Office of the Inspector General released a 98-page audit of its Flock system in July 2026. It found that 161 searches in a two-month period generated false stolen-vehicle alerts, a 32.3% false positive rate, each resulting in a stop of an innocent driver. The IG characterized the errors as creating “individual liberty interests” and “substantial legal and financial liability concerns.” LAPD let its Flock contract expire. Troy’s audit reports no alert accuracy data.
In Fargo, North Dakota, the police department launched a public Flock transparency portal in July 2026. Audit logs are accessible online, showing the time, date, purpose, and officer identifier for each search. In Binghamton, New York, under the same state law as Troy, the City Council is currently reviewing its Flock contracts. The proposed retention policy is 30 days. More than 80 localities across 28 states have terminated or deactivated Flock contracts since 2025, citing data-sharing violations, unauthorized federal access, and governance failures.
The Council’s proposed 48-hour deletion window is stricter than the 30-day standard being debated in other jurisdictions. The audit data supports it.
What the Council asked for and what it received
The May 19 agreement required the Administration to provide data to the Council during the 60-day period. The period closed July 18. As of July 24, the Council President had not seen the results. The audit existed. It was not delivered until July 29, eleven days after the deadline.
Troy pays $78,000 per year for 26 Flock cameras. The Council’s oversight role over that contract is the reason the May 19 agreement existed. The agreement required delivery by July 18. The Council received the audit eleven days later.
At the very least, the audit provides no evidence for a retention window longer than 48 hours.
More than 80 municipalities across 28 states have cut ties with Flock since 2025. The data from Troy explains why.
Source documents: 60-Day Narrative (PDF). Search logs on file.