Cybersecurity
What Boards Should Know About Today’s Threat Actors
Governance Research
Governance Surveys
Directorship Magazine
Key Points
- Boards should recognize that AI failures in critical infrastructure are often irreversible, requiring a shift to physical consequence modeling.
- Standard IT reviews fail to account for the unpredictability of physical operations, making staged deployments and rigorous manual fallback testing essential.
- Organizations should assign a single owner to manage the accountability gap between IT regulations and operational technology safety standards to ensure full regulatory compliance.
This AI-generated summary, based on content on this page, was reviewed by NACD editors for accuracy.
Standard AI governance fails when it comes to critical infrastructure. These are the four questions boards should ask to oversee physical AI risks.
A technology committee approves an artificial intelligence deployment for predictive maintenance at a generation facility. The proposal passed every standard enterprise AI review. The model is validated, the vendor is credible, and the risk assessment is signed off.
Six months later, the model begins to drift and miss maintenance alerts. A turbine runs past its service threshold. The failure is traced back to a shift in sensor data patterns the model had never seen in training.
Nobody tested for that possibility. Nobody asked: What happens when this model is wrong in a physical environment?
This is already happening. Every time a board overseeing an energy, utility, or industrial company approves an AI deployment ...
Thank you for your interest in this page.
Member-Only Content
For full access, please log in, or explore membership options.
Alexander Antukh, NACD.DC®, is chief information security officer of AboitizPower, the Philippines’ largest energy producer, where he oversees AI, cybersecurity, and operational technology security governance across dozens of generation facilities. He is a board member of the FAIR Institute and the founder of Cyber Hermes, a board-level AI and cybersecurity advisory firm.
