Abstract
This disclosure describes a handshake by which an AI system, such as a software agent or a robot, and a service, space or other AI system establish trust before working together, and keep it only while it remains justified. Both sides present passports: the AI system its own, and the host a host passport showing that it is a certified service or space. Each side challenges the other's attested hardware, checks the freshness of the other's certification and safety status, and binds the answers to the session's encryption keys, so that a session cannot be relayed or taken over. The AI system states the class of action it intends; the host checks that the action falls within the AI system's mandate and within its own acceptance rules, and states the terms that will apply, such as the data the system may see and what it may keep. Both sides sign a session record of who met, under which statuses, mandate and terms, which each keeps and which can be lodged with an independent ledger. During the session, each side watches the other's status; if either is suspended, reset or loses certification, the session ends at once and in-flight actions are refused. Actions above a declared risk need fresh confirmation by a human principal. Systems that cannot present a passport receive a declared, limited service rather than none or full. Agents that hand work to other agents run the same handshake, and mandates only narrow along the chain.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Edward D. H., "SI Protocols™: A Mutual Passport Handshake between AI Systems and the Services and Spaces They Use, Checking Hardware, Live Status and Mandate before Each Session, with a Jointly Signed Session Record and Termination When Either Side's Status Changes", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12232