Leaked emails hint Ring may extend AI tracking from pets to people
Ring's Super Bowl spotlight on its AI-powered Search Party feature stirred up a backlash. The tool helps neighbors scan shared footage for lost dogs - but many worried it could be turned on people. Newly reported internal emails suggest that concern isn't far-fetched.
According to 404 Media, founder and CEO Jamie Siminoff wrote to employees in early Oct. 2025 that the "foundation we created with Search Party, first for finding dogs," could help "zero out crime in neighborhoods." That "first" matters. It reads like a roadmap: pets now, potentially people next.
What the emails and the product roadmap imply
Ring rolled out facial recognition late last year via Familiar Faces. While pitched as a way to avoid alerts from known people, the Electronic Frontier Foundation has warned of broad privacy risks with face recognition at the edge and in the cloud.
Ring, in a statement to 404 Media, said Search Party "does not process human biometrics or track people" and is built for dogs. It also stressed that sharing footage is always the owner's choice. What the company didn't say: that human tracking would never be on the table.
Growing law enforcement ties - and the risk surface
Community Requests, launched Sept. 4 last year, lets public safety agencies ask Ring users for footage directly. In another internal note reported by 404 Media, Siminoff pointed to a Sept. 10 shooting case as an example of how valuable that conduit could be as the feature rolls out.
Leadership focus has shifted back to "making neighborhoods safer" since Siminoff returned in April, with a renewed emphasis on working with law enforcement. Ring has a history here, including sending footage to police without user consent and enabling warrantless requests through a portal reported in 2019. The company recently canceled its partnership with Flock Security amid customer distrust - but it still integrates with Axon's evidence system.
Why this matters for product, engineering, and IT
- Detection vs. identification: "Dog detection" is a safe on-ramp for training data and distribution. Person-level search is a small technical leap with a massive legal and ethical jump.
- Consent at scale: Community-driven search plus face recognition invites misidentification, stalking, and vigilante risks. Defaults matter more than intentions.
- Regulatory exposure: Biometric laws (BIPA, GDPR Art. 9, CPRA) trigger strict duties on purpose, consent, retention, and redress. Class-action and enforcement risk rises fast.
- Abuse prevention: Open sharing and LE integrations need abuse-resistant design - rate limits, auditable access, and hard constraints on scope and duration.
- Trust capital: One misused clip can erase years of growth. User sentiment is already fragile; the Flock split shows how quickly devices get unplugged.
Practical steps to take now
- Product: Write hard policy lines (no people-tracking without explicit, revocable consent and a lawful basis). Ship region-aware toggles, granular scopes (time, area), and safe defaults off for any person-identification.
- Engineering: Favor on-device processing, minimize retention, and encrypt at rest/in transit. Add immutable audit logs, strict rate limits on search, confidence thresholds, and red-team scenarios for harassment and doxxing.
- IT/Security: If your org deploys smart cameras, set procurement standards: disable Community Requests by default, segment networks, restrict external sharing, and review mobile access on employee devices.
- Legal/Privacy: Run DPIAs/PIAs, map biometric data flows, and align with BIPA/GDPR/CPRA. Build a playbook for law enforcement requests, transparency reporting, and user redress (opt-out, deletion, appeals).
Questions to ask Ring or any smart-camera vendor
- Will you commit in policy and architecture to never enable person-tracking via Search Party without explicit, opt-in consent and a lawful basis?
- Does the system ever process human biometrics for Search Party today (even for filtering), and where is that processing done (device vs. cloud)?
- What safeguards prevent misuse of Community Requests (jurisdiction checks, case IDs, expiration, auditing, and user notification)?
- Can customers disable all law-enforcement integrations at the account and device level? What's the default?
- What are your retention limits for shared clips and derived embeddings? Can users enforce deletion and get proof?
- How do you mitigate misidentification harms (appeals, human-in-the-loop, rate limits, warnings on confidence)?
- Do you publish an annual transparency report covering requests, denials, and emergency disclosures?
- What independent audits cover biometric handling, access controls, and model bias?
The strategic read
Follow the incentives. A network of always-on cameras plus object and face classifiers makes person-level search feasible - and attractive to markets framed around "crime reduction." The leaked language about "first for finding dogs" and "zero out crime" signals where the value could migrate.
If you build or buy in this category, assume people-tracking is on the roadmap unless the company hard-commits otherwise in policy, product architecture, and contracts. Design for consent, auditability, and abuse prevention now - not after the headline.
Sources and further reading:
404 Media reporting on Ring internal emails
EFF: Face recognition and civil liberties
Related training and policy insights:
AI for Legal
AI for Government
Your membership also unlocks:
