Abstract
This disclosure describes an identification system for the parts of the AI world: AI systems, models, robots and devices, sensors, governed software bundles, certified spaces and the organizations responsible for them. Each receives a lifelong identifier from one registry, under a common scheme with a type prefix, a check value and a public key. Identifiers are never reused, survive changes of ownership, and remain resolvable if the maker fails, because records are held by the registry and mirrored by independent parties. Records are linked by typed, signed relations: a model derived from another model, a bundle containing a model, a robot running a bundle, a sensor fitted to a robot, a robot operating in a space, a system accountable to an organization. The relations form a graph that changes over time and keeps its history. When a flaw is found in one item, such as a defect in a sensor model, a vulnerability in a bundle or a harmful behavior traced to a model's training, the registry follows the graph to find every system that contains, runs or derives from it, at any depth, and marks each affected record, so that the certification status of each affected system can be changed and its operators told. Each record has public fields, available to anyone, and restricted fields, available to authorized parties, so that tracing does not expose personal or commercial data.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Edward D. H., "SI ID System™: Lifelong Identifiers for AI Systems, Models, Robots, Sensors and Software Bundles in One Registry, Linked as a Graph so that a Flaw Found in One Can Be Traced to Every System It Affects", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12243