Abstract
This disclosure describes an island mode for the AI systems of an organization or a single device. All traffic between the organization's governed AI systems and the outside world passes through one governed gate. When the gate closes, whether for maintenance, on detection of an attack, on loss of outside connectivity, or by order of an authorized person, the organization's own AI compute, governed data custody and personal AI devices keep serving each other inside, under locally held, signed policies and credentials with declared validity periods. While the island lasts, governance tightens as a declared function of time offline: the runtime governor raises its level step by step as policies, revocation lists and knowledge age, so a long disconnection makes the systems more cautious, never less. The gate reopens only on the release of two authorized people and in stages: first a check that nothing inside was compromised, then admission of signed, verified updates inbound only, then reconciliation of the local tamper-evident record with outside registries, and finally a stepwise return to normal governor levels. Island mode is drilled on a declared schedule and the results support certification. The method lets an organization keep its intelligence running through outages and attacks while staying safe when cut off.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Edward D. H., "OfflineSI™: Island Mode for an Organization's AI Systems, with Governance that Tightens with Time Offline and Staged Reopening", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/12154