Gartner forecasts that by 2028, misconfigured artificial intelligence within cyber-physical systems will cause a national critical infrastructure shutdown in a G20 country. Cyber-physical systems encompass engineered setups integrating sensing, computation, control, networking, and analytics to engage with the physical world, including operational technology, industrial control systems, industrial Internet of Things, robots, drones, and Industrie 4.0 technologies.
Risks Stem from Engineering Errors, Not Attacks
Gartner VP Analyst Wam Voster warns that future infrastructure failures may arise from well-intentioned engineers, flawed update scripts, or minor errors like misplaced decimals, rather than cyberattacks or natural disasters. Misconfigured AI can autonomously halt vital services, misread sensor inputs, or execute unsafe commands, resulting in physical damage, widespread service outages, and threats to public safety or economic stability. Power grids exemplify the vulnerability: AI models balancing supply and demand might erroneously detect instability, triggering unnecessary isolations or load shedding across regions or nations.
Voster describes modern AI models as “black boxes” where even developers struggle to anticipate outcomes from configuration tweaks, amplifying risks as systems grow more opaque.
Mitigation Demands Human Overrides and Testing
Gartner urges chief information security officers to implement secure “kill-switch” or override modes in critical infrastructure cyber-physical systems, accessible solely to authorised operators to ensure human control overrides full autonomy. Organisations must develop full-scale digital twins for pre-deployment testing of updates and configurations, alongside real-time monitoring with rollback capabilities for AI changes. National AI incident response teams should coordinate responses to misconfiguration events.
Cyber-Physical Systems Heighten Stakes
Cyber-physical systems demand rigorous safeguards given their integration of IT and physical operations. AI adoption in cybersecurity and operational technology leads at 57 percent, yet cybersecurity ranks as the top scaling barrier at 23 percent, with data challenges following closely due to quality, access, lineage, and governance issues. Gartner clients can access the full Predicts 2026 report on emergent AI risks in cyber-physical security.
