Deployments
Ambient agents run continuously and act on what they sense without a request
Ring's Familiar Faces and Gemini for Home both analyze home camera video around the clock. Recall builds a searchable history of everything shown on a PC screen. Scale figures for ambient clinical scribes and retail vision systems were not verifiable for this edition.
| Deployment | Developer / operator | Where | Scale | Autonomy | Status | Controls declared |
|---|
Recent events
- Google planned Gemini camera features for 13 more European and Japanese markets in early 2026.
- Contractors in Kenya reviewed in-home footage from Meta glasses, Swedish papers reported in February 2026.
- Alexa+ can work with Ring cameras to flag unusual activity.
Attack surface
Two of ten attack-surface dimensions rate very significant or critical
Data capture rates critical. Ratings use five levels, from limited to critical, and cover the paths an attacker or a failure can take into the agent.
| Dimension | Rating | Why |
|---|---|---|
| Physical access | Significant | Sensors sit in shops, clinics and offices that many people pass through |
| Identity and authentication | Moderate | Device and operator credentials, often vendor-managed |
| Command and control channels | Moderate | Few sensors take commands beyond configuration |
| Perception and sensors | Very significant | Cameras, microphones, palm and Wi-Fi sensing are the product |
| Connectivity surface | Significant | Continuous uploads to vendor clouds |
| OTA and update pipeline | Significant | Models and firmware update remotely |
| Data capture and retention | Critical | Capture of people who never agreed to it is the main exposure |
| Integrations and permissions | Significant | Feeds go to EHRs, loyalty programs and HR systems |
| Behavioral and policy boundary | Moderate | Retention and use limits are set by policy, not by the device |
| Multi-agent coordination | Moderate | Sensor networks aggregate across sites |
Incidents
No registry incident involves ambient sensors yet
The 137AI Incident Registry logs an incident when an agent's failure or misuse caused harm or a regulator acted, and public sources name the developer, the failed control and how it was found. No ambient sensors case meets that bar yet. Candidates go on the review list in the registry methodology.
Rules
Emotion recognition is banned at work in the EU; elsewhere only face rules apply
| Rule | Jurisdiction | What it requires | Status |
|---|---|---|---|
| EU AI Act Article 5(1)(f) | EU | Bans emotion recognition in workplaces and schools, except medical or safety use | In force Feb 2, 2025 |
| CAC and MPS face recognition measures | China | Notice and consent; bans in hotel rooms and restrooms | Effective Jun 1, 2025 |
| Home Office biometrics consultation | UK | Asks whether rules should cover gait, voice and emotion inference | Closed Feb 12, 2026 |
Status colors: green in force, blue voluntary standard or framework, amber adopted and phasing in, gray draft or pending, red withdrawn.
Record
Page Record
- Agent class
- Personal and ambient
- Datasets
- ai_deployments.csv (2 rows for this page); ai_attack_surface.csv (10 rows); ai_incidents.csv (0 linked)
- Key sources
- EU prohibited practices (Orrick)
China face rules (China Briefing) - Last verified
- October 2026