A Retreat, Not a Removal
On July 10, 2026, Flock Safety published a blog post ending "Distress Detection," a feature that used the always-on microphones bundled with its acoustic gunshot sensors to flag human screaming and route an alert to police. The Electronic Frontier Foundation, which had campaigned against the feature since it launched in October 2025 under the product name Raven, declared the reversal a "Victory" on July 17. But EFF's own account is careful to note what didn't change: the microphones stay mounted on poles above city streets, still listening for gunshots, and Flock can re-enable voice-distress analysis with a software update whenever it chooses.
That distinction — a capability withdrawn versus a capability that never left the hardware — is the real story here, and it says more about how surveillance policy is failing than the pilot's cancellation says about how it's succeeding.
The Case Flock and Police Departments Would Make
It's worth taking the pitch seriously before dismissing it. Gunshot-detection microphones exist because 911 calls undercount shootings — many go unreported, especially in neighborhoods where residents distrust police or simply don't hear a single shot as unusual. Extending that same sensor network to detect screaming is, on its face, a plausible public-safety upgrade: a domestic-violence victim or assault target who can't reach a phone might still be heard. Cities that adopted Flock's network of over 6,000 communities did so because license-plate and gunshot data has produced real investigative leads. Faster alerting to distress sounds, in isolation, is a defensible extension of that logic, not an obviously reckless one.
Why the Steelman Still Fails
The problem is that "in isolation" is not how this technology was deployed. Distress Detection went live as a firmware update to hardware that was already installed under a different stated purpose — gunshot detection — without new community consultation, a new procurement vote, or new legal review of the expanded audio-processing capability. EFF's central legal objection is that ambient, always-on interpretation of human speech runs into state all-party-consent eavesdropping statutes that were written decades before software could parse a scream out of street noise, and whose application to this use case remains genuinely unresolved. Flock's decision to retire the feature by blog post — rather than by removing the audio-analysis capability from the devices — means the same unresolved legal question can resurface at any city's request, or Flock's, with no public process at all.
This is precisely the "mission creep" dynamic EFF has documented across Flock's product line. The company's own scale illustrates the stakes: Flock now operates more than 80,000 camera and sensor devices nationwide for a company valued near $7.5 billion, and a recent LAPD audit found the system had wrongly flagged 161 people's vehicles as stolen — a reminder that scale amplifies both false positives and the consequences of quietly expanded capabilities.
The Regulatory Response Is Starting to Get the Model Right
Encouragingly, the policy response taking shape in 2026 isn't blanket prohibition — it's purpose limitation with enforceable teeth, which is the proportionate approach this publication has consistently favored over outright bans on useful public-safety tools. Washington's Driver Privacy Act (SB 6002), signed by Governor Bob Ferguson on March 30, 2026, doesn't ban automated license-plate readers; it caps data retention at 21 days, bars use for immigration enforcement or tracking reproductive-healthcare visits, and requires agencies to register systems with the Attorney General. At the federal level, Rep. Keith Self's PRIVACY Act, introduced July 15, 2026, would require federal agencies to get a warrant before querying state and local surveillance data, including Flock's network — again, a gate on use, not a ban on the sensors themselves.
That's the right template, and it's exactly what's missing for audio capability toggles specifically. A city that contracts for gunshot detection should be able to trust that "distress detection is off" means the underlying voice-processing capability has been disabled at the hardware or firmware level in a verifiable way — not that a vendor has flipped a switch it can flip back. Procurement contracts should specify which audio-processing features are authorized, require a new council vote to reactivate any retired one, and mandate audit logs of every feature-level software update pushed to deployed sensors, the same way health and safety inspectors log changes to regulated equipment.
The Real Lesson
Flock's reversal shows that public pressure and EFF-style advocacy can still move a surveillance vendor without waiting for legislation — that's a genuine win worth crediting. But it also shows the limits of relying on vendor discretion: the company that added Distress Detection without a public process is the same company deciding, unilaterally, to remove it. Lawmakers extending SB 6002-style purpose limitations and warrant gates to software feature changes on already-deployed sensor hardware — not just to new device purchases — would close that gap without slowing the deployment of gunshot detection that genuinely helps police respond faster to real emergencies.