Abstract
This disclosure describes a method of governing the conclusions that AI systems draw about people, as distinct from the data they draw them from. An AI agent that watches a person's messages, screen, purchases, location and wearable sensors can infer things the person never stated, such as a likely illness, a change in mood, money trouble, a pregnancy or a new relationship, and can act on, show, share or store those conclusions. In the method, every such inference about a person is recorded as a derived record with its class, its confidence, the source records it was drawn from and the person it concerns. Each class of inference is governed by an inference contract chosen by that person, stating whether the system may draw it at all, keep it, show it, act on it, or share it, with whom, for how long, and with what wording. Sensitive classes are off by default. The governance point refuses actions that rely on an inference beyond its contract, and the egress gateway blocks outputs that carry it. Inferences about people other than the owner, such as visitors or colleagues seen in screen history, are not kept or used by default. The person can see every inference held about them, correct it or delete it, and deletion propagates to everything derived from it. Health and similar inferences are presented as patterns to discuss with a professional, never as diagnoses.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Edward D. H., "SI Data Bank™: Inference Contracts that Govern What AI Systems Conclude about People, Enforced at the Governance Point and the Egress Gateway", Technical Disclosure Commons, (October 13, 2026)
https://www.tdcommons.org/dpubs_series/12173