Flock's new surveillance guardrails skip the cameras already in the field

Kenneth Fitzgerald, the police chief in Bellingham, Massachusetts, found out what his cameras could do in the middle of a criminal investigation. In a public letter he said the department learned that Flock "Free Form", an AI tool that searches video using plain-language queries, was available through cameras it already had, and that he does not believe Flock clearly explained or identified those features to the department ^[8][9]. He had told his own public the system read license plates and nothing else ^[10]. That gap is about what the buyer knew he had bought, and it closes with a contract and a change notice rather than a settings screen.

The renewal numbers are the part Flock can see from its own side. Secure Justice counts 214 cities and counties dropping Flock since 2021 ^[4], and 90 of those cancellations fell in August 2026, four times the month before ^[5]. Techdirt's updated look at the same chart adds 10 more and puts August at 96 ^[6]. That single month is 44.9 percent of the entire series, 96 divided by 214 ^[11], and working back from the fourfold figure, July sat around 22 ^[12]. Cyrus Farivar reported the underlying data for Ars Technica in a piece first published on August 28 ^[7]. The reasons cities gave include high expenses and unwanted data sharing alongside the abuse reports ^[13], which is the composition of a procurement complaint rather than a press cycle.

Against that, the guardrails: shorter retention by default, more information demanded of an officer running a search, and a misuse flag routed back to the department ^[1]. They are reasonable things to ship, and narrower than they sound. Techdirt reports the new restrictions do not appear to reach systems already in use and can be switched off by the agency ^[3]. The accounts that hold these controls, in other words, are largely the accounts that have not signed yet.

The renewal losses, more than the coverage, are what Techdirt credits for the change ^[2], noting that for the previous couple of years Flock answered critics by calling the reporting misleading or by suggesting the critics opposed public safety ^[14]. The material carries no ship date for the new settings, so the sequence is an inference rather than a documented before-and-after.

A vendor's answer here comes down to two questions: whether the control reaches equipment already installed, and whether the local agency can turn it off by itself. The weakest of the four cells is a control that arrives on by default, can be reversed locally, and never touches installed hardware, and that is the cell described in this reporting ^[3]. The forcing function for a buyer is duller than a settings audit: a written list of every capability the account is entitled to run today, plus a clause requiring notice before that list changes. Fitzgerald's other complaint, that Flock tried to modify some terms of service and access while the contract was still in place ^[15], is the same gap viewed from the vendor's end. A department that cannot produce that list without calling its rep cannot make the statement Fitzgerald made to his town, and it will be the one issuing the correction.