Abstract
This disclosure describes a way of making human accountability for AI systems and robots operational rather than nominal. Every certified AI system or robot is bound, in a public registry, to an SI Manager: a person holding a current credential for the system's safety profile, together with at least one registered deputy. The binding is not a line in a document; the system's own runtime checks it. At start, the system verifies that a current manager binding exists before unlocking its capabilities; while running, it verifies that the binding remains current. If no manager is bound, the manager's credential has lapsed, or neither the manager nor a deputy has acknowledged within the required time, the system steps down to a declared unmanaged profile, a reduced set of capabilities defined in advance for that system, and its public status shows "unmanaged" until a manager is bound again. Reachability is proven by signed acknowledgements of the system's periodic status and of its alerts, within a response time set by the system's risk profile; an unacknowledged alert escalates to the deputy and then triggers the step-down. Competence credentials expire and are renewed only by recorded training and drills. Each manager's span of control is capped by the summed risk weight of the systems bound to them, and duty-time limits apply to the highest risk profiles, so that accountability cannot be spread across more systems than a person can actually watch. Responsibility passes from one manager to another only by acknowledged handover, the incoming manager confirming before the outgoing manager is released. The manager holds keys that can always stop, restore or tighten the system's limits without depending on the registry, but cannot loosen limits alone. Managers may not certify or inspect systems they manage. The public sees that each system has a current manager and how it is covered, while the manager's personal identity is disclosed only to those with standing. A coordination layer serves the manager as a tool: it routes alerts to the manager and then the deputy, records signed acknowledgements, and may change a system in its operations lane only while the system has both a current manager and a current digital twin; larger changes need the affected managers' acknowledgement of the twins' recorded predictions. An organisation may name a senior person accountable for all its managers, and credentialed auditors of accredited bodies audit bindings, acknowledgements and drills.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Edward D. H., "SI Manager™: A Credentialed, Accountable Person Bound to Each Certified AI System or Robot through a Public Registry, in which the System Checks for a Current Manager at Start and While Running and Steps Down to a Declared Unmanaged Profile When None Is Present, Reachability Is Proven by Signed Acknowledgements within a Response Time Set by Risk, Competence Credentials Expire, Each Manager's Span of Control Is Capped by Risk Weight, Responsibility Passes Only by Acknowledged Handover, and the Manager Can Always Stop or Tighten but Never Loosen Alone", Technical Disclosure Commons, (November 10, 2026)
https://www.tdcommons.org/dpubs_series/12957