Abstract
This disclosure describes a method by which people who are not the owners of an ambient AI system, such as guests, family members, colleagues or passers-by, can tell every certified sensor and AI agent around them how they wish to be treated, without having to find, ask or trust each device's owner. A person's phone, watch or a small tag broadcasts a short-range privacy signal stating a preference level, such as "do not process me", "presence only" or "events about me, no inferences". The signal uses rotating, unlinkable identifiers, so that it cannot itself be used to track the person, and is signed so that it cannot be forged to switch off someone else's protection. Certified sensors honour the signal at the source, before anything leaves the sensor, by suppressing, masking or reducing what concerns that person; certified agents honour it by drawing no inferences about that person and taking no action concerning them. Where a sensor cannot reliably separate the signalling person from others, it reduces its output for everyone present. Each device records, without identifying the person, that a signal was received and how it was honoured, so that compliance can be audited and tested for certification. A host may also switch a whole space into guest mode, which applies the strictest default to everyone not enrolled.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Edward D. H., "SI Sensor™: Bystander Privacy Signals for Ambient AI, Honoured by Certified Sensors and Agents at the Source, with Unlinkable Signals and Verifiable Compliance", Technical Disclosure Commons, (October 13, 2026)
https://www.tdcommons.org/dpubs_series/12192