Abstract
This disclosure describes a way to test a whole sector's readiness for AI-driven attacks by live drills rather than discussion alone. Each participating institution, such as a bank, payment operator, exchange, utility or hospital group, provides a digital twin of its relevant systems: copies of its certified AI systems and software bundles, synthetic data with the statistical shape of its real data, and models of its essential services and their minimum capacities. The twins are connected inside certified containment facilities into a model of the sector, including the links between institutions. A drill injects a scripted AI-driven attack across the twins at once, such as agents acting outside their mandates, a compromised shared model, attacks on data custody at machine pace, or a worm pooling devices. The institutions respond with their real procedures and tools: governors, authority leases, island mode, quorum signing for fleet-wide reset, staged restore and fallback for essential services. Every step is timed and recorded. Results are measured against declared objectives, such as time to contain, time to restore, and whether essential services kept their minimum capacity. Findings update the pre-signed playbooks used for fleet-wide resets, become tests in the standard through its emergency track where urgent, and are reported to participants in full and to the public as summaries without methods or institution names. Drills recur, and institutions' results contribute to their certification.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Wang, Edward D. H., "SI Drill™: Live Sector-Wide Drills against AI-Driven Attacks, Run on Digital Twins of Many Institutions inside Certified Containment Facilities, Testing Reset, Isolation and Recovery End to End, with Results that Update Signed Playbooks and the Standard", Technical Disclosure Commons, (November 02, 2026)
https://www.tdcommons.org/dpubs_series/12615