Sources
All accessed 2026-08-19. [vendor] marks vendor-published claims about the publisher's own market.
NERC, AI and ML in Real-Time System Operations, white paper rev 1 (Nov 2024): the operator-authority statement, the n=47 adoption survey, the alarm-triage and model-reconciliation opportunities, and the "humans as hazards" caution; NREL/TP-5D00-87440, eGridGPT: Trustworthy AI in the Control Room (May 2024): the validation loop, local open-weight deployment for compliance, and the model-conversion result validated against the model-validation standard; CISA with ASD ACSC, NSA, FBI, CCCS, BSI, and the Netherlands, New Zealand and UK cyber centres, Principles for the Secure Integration of AI in Operational Technology (Dec 3 2025) [verbatim wording to verify against the primary PDF]; EEMUA 191 Edition 4 (Nov 2024) via the IChemE Hazards 35 update paper (2025), with IEC 62682:2022 alignment; CSB final report on the BP-Husky Toledo incident (Jun 2024); Ofgem, Ethical AI use in the energy sector v2 (May 13 2026) including the agentic-AI framing, human-oversight practice, and skill-base failure mode; Yokogawa and ENEOS Materials (adopted Mar 2023) and Yokogawa and Aramco Fadhili (Oct 2025) via trade press [vendor releases to verify directly]; Honeywell with Chevron (Oct 1 2024) and TotalEnergies Port Arthur (Nov 2025) [vendor]; BuildSafe (Frontiers in Built Environment, Jun 18 2026); ISO 55013:2024 with ISO 55001:2024 clause 7.6; EPRI Open Power AI Consortium coverage (Apr 2025); FoSMo award coverage (Jun 2026); Argonne GridMind (SC'25 workshops, Nov 2025); CERN UNICOS operator assistant (arXiv, Sep 2025); Ofwat AI adoption plan (Jun 2026); DESNZ independent review terms of reference (Dec 2025, updated Mar 2026). Re-verify by Q4 2026: the multinational OT guidance verbatim; Yokogawa primary releases; SAP asset-agent names; the DESNZ review if published; CIP-015 dates against the primary rule.
Source: research/R05-lob-and-ot/sources.md in the evidence repository behind this site.