Amit Pawar, SVP, Solutions Advisory & Customer Success, Xage Security

Artificial intelligence is no longer a future concept in operational technology (OT). Across critical infrastructure, AI is rapidly moving from pilot initiatives into production environments. It is improving maintenance efficiency, accelerating diagnostics, and supporting real-time operational decisions.

But AI adoption in OT is fundamentally different from traditional IT modernization. AI introduces new safety, reliability, and security risks into environments engineered for determinism and containment. Rogue AI, opaque decision logic, unsafe recommendations, and expanded cyber attack paths can undermine operational resilience if left ungoverned.

This is why the CISA-led joint guidance, “Principles for the Secure Integration of Artificial Intelligence in Operational Technology,” published December 3, 2025, is such a pivotal milestone for the industry.

The guidance provides a practical framework written explicitly for critical infrastructure owners and operators. It raises an urgent question for OT security leaders: How do you translate AI principles into controls you can actually enforce in live OT environments?

The Four Principles and the Execution Gap

CISA’s guidance organizes secure AI-in-OT adoption around four core principles:

• Understand AI
• Consider AI Use in the OT Domain
• Establish AI Governance and Assurance Frameworks
• Embed Safety and Security Practices into AI and AI-Enabled OT Systems

While the principles are clear, execution is not.

OT environments operate under constraints most IT security models fail to account for. These include legacy systems with long lifecycles, vendor-dependent access, intermittent or disrupted connectivity, and an uncompromising priority on safety and availability. In this reality, policy documents alone are insufficient.

What is required is an architecture that can continuously verify, constrain, and audit AI behavior, even when systems are decades old and networks are degraded.

Why Zero Trust Enforcement Is Essential

CISA’s message is unambiguous. AI can deliver real operational value to critical infrastructure, but only if it is deployed with rigorous safety, security, and governance. That means treating AI not as a trusted background process, but as a privileged participant in OT. It must be assigned a distinct identity, granted only tightly scoped permissions, continuously monitored and auditable, and governed with human oversight.

This is where Zero Trust enforcement becomes essential.

The Xage paper, “Operationalizing CISA’s AI-in-OT Principles: Zero Trust Enforcement with Xage,” translates the four principles into enforceable implementation patterns that OT teams can deploy today, despite legacy constraints and real-world operational conditions.

Xage was built specifically for this challenge. As a Zero Trust platform designed for critical infrastructure, Xage aligns directly with what the guidance demands. This includes policy-enforced access, containment by default, audit-ready operations, and resilience in disconnected or degraded environments.

Read the full paper, “Operationalizing CISA’s AI-in-OT Principles: Zero Trust Enforcement with Xage,” to see how guidance becomes enforceable OT security controls.