Key facts
- The Operational Technology Cybersecurity Coalition (OTCC) released a report titled 'The Missing Pillar of AI Security: Critical Infrastructure Resilience' with 18 recommendations for U.S. policymakers.
- Four summer roundtables drew on nearly 70 representatives from government, the cybersecurity community, and critical infrastructure sectors to shape the OTCC recommendations.
- OTCC's recommendations aim to improve resilience during disruptions, strengthen baseline cybersecurity, and deploy AI-supported defenses inside OT environments.
- The coalition calls for money to go to the Cyber Response and Recovery Fund and for the Cyber Safety Review Board to be reconstituted.
The Operational Technology Cybersecurity Coalition released a report titled The Missing Pillar of AI Security: Critical Infrastructure Resilience with 18 recommendations for U.S. policymakers.
The coalition said critical infrastructure resilience should be a central pillar of national AI security policy. It also said owners and operators of critical infrastructure, along with the broader U.S. government, should apply zero trust principles, a security model that assumes compromise and verifies every request, to AI models and agents, embrace CISA’s CI Fortify approach, and assume their own networks or networks they depend on have already been compromised.
Four summer roundtables drew on nearly 70 representatives from government, the cybersecurity community, and critical infrastructure sectors to shape the recommendations. The coalition’s goals include improving resilience during disruptions, strengthening baseline cybersecurity, and deploying AI-supported defenses inside operational technology environments.
The coalition wants regulators to close gaps across sectors without copying requirements, fund under-resourced entities, create an AI governance framework for operational technology, and promise lasting access to AI-powered cybersecurity tools. It also said funding should support the Cyber Response and Recovery Fund and that the Cyber Safety Review Board should be reconstituted.
OTCC said the priority of sustaining and restoring essential services is echoed in U.S. strategies such as NSM-22 and the 2026 National Resilience Strategy. Recent U.K. guidance also echoes this priority, the coalition said.
Policy director Michael Garcia said Congress and the Administration do not need to build everything anew because each recommendation relies on existing authorities, programs, and institutions. OTCC said success for national AI security policy should be judged by whether operators of critical infrastructure can keep essential services running and restore them quickly after an incident. Roughly 70% of operational technology vulnerabilities reside within legacy or deeply embedded systems, and patching or rebooting them would halt nonstop physical operations.
CISA monitors more than 12,000 industrial control system vulnerabilities, and a shortage of specialized engineering talent creates serious bottlenecks in testing them. In August, CISA and federal partners warned that adversaries use AI-generated scripts to identify and exploit longstanding weaknesses in factory controllers spanning the power, water, and manufacturing industries at faster speed and broader scale.
The coalition said AI firms should work with operational technology and federal stakeholders to prepare models using sensitive infrastructure data under safe conditions. One example it gave was holding protected utility data within owner-controlled spaces or closed test beds.
OTCC plans to issue additional studies that look more closely at urgent topics. The coalition said the chance to strengthen defenses before the next major AI-enabled cyber attack is narrowing but remains open.
A nationwide FAA ground stop and a Microsoft outage last week caused nationwide chaos and brief fears of a coordinated cyber attack, though neither incident was actually a cyber attack. OTCC linked the FAA ground stop and Microsoft outage to the size of the disruption attackers want to create, now with added speed and sophistication from AI. Malicious cyber actors have breached operational technology systems using weak passwords, missing multifactor authentication, excessive access rights, and poor network segmentation, and AI will make these weaknesses worse.
Organizations deploy AI agents and other machine-based identities faster than they can govern them, with these identities potentially exceeding human staff by more than 100 to 1. A Nozomi Networks survey found that fewer than 8% of organizations have put AI into use across multiple operational technology cybersecurity functions.
OTCC urges Congress and the Executive Branch to create a national approach based on CISA’s Cross-Sector Cybersecurity Performance Goals, with mandatory harmonized requirements, secure-by-design mandates for producers, liability protections, and financial help for under-resourced entities. The framework must account for autonomous software agents and other non-human identities that can reach sensitive systems.
The coalition recommends an operational technology-specific AI governance framework, scaling state and regional models, ensuring long-term access to AI-enabled capabilities, accelerating adoption of certified commercial off-the-shelf products, and testing autonomous AI deployments. OTCC backs renewed backing for information-sharing and threat-analysis centers, a public-private forum connecting AI developers and operational technology teams, and tighter international coordination.

OTCC argues federal cybersecurity policy should stress withstanding, containing, and recovering from a successful attack because strong controls cannot remove every vulnerability. It recommends that critical infrastructure owners and operators regularly assess recovery readiness, including detection across operational assets, operational technology-specific response coordination, and restoration of priority functions with minimal disruption.
The coalition intends to hold meetings with policymakers, AI developers, and the operational technology community in the coming months to test and refine the ideas. It also plans to reach out to Congress, federal, state, and local governments and AI labs to push implementation.
OTCC notes CISA faces low staffing and insufficient resources while implementing several measures, and says rebuilding and supporting CISA and sector risk management agencies is imperative.
OTCC pressed CISA to establish mandatory security requirements for operational technology across federal civilian agencies. In an Oct. 6 report, the coalition called for a required security order for federal operational technology, saying no current rule sets baseline practices for federal operational technology and CISA lacks visibility into the risks.
Federal agencies rely on operational technology in more than 8000 General Services Administration-managed facilities, including laboratories, hospitals, and ports of entry, for HVAC, power, access control, water, and building automation systems.
A Sept. 30 Government Accountability Office report found only seven of 22 civilian agencies had fully met Office of Management and Budget rules to list networked operational technology and Internet of Things devices. The inventory deadline was September 2024, and the Office of Management and Budget had not released updated fiscal year 2026 guidance.
The proposed directive would make agencies appoint a senior official or office to run operational technology security and fold operational technology risk into enterprise risk management. It would set a baseline covering verified recovery, incident preparedness, configuration management, remote access, network segmentation, and asset inventory.
OTCC’s priority controls include changing default passwords, multifactor authentication, segmentation, and backups. The coalition asked CISA to emphasize these measures together with Office of Management and Budget requirement oversight, although the report does not request patching or firmware updates.
John Gallagher, vice president at Viakoo, said an inventory alone is not enough. He warned that missing remediation could create backlogs overwhelming operational teams without automated patch and configuration management.
Louis Eichenbaum, federal CTO at ColorTokens, said containment matters since patching many industrial devices quickly would interrupt operations.
OTCC said the directive would support CISA’s CI Fortify resilience initiative by establishing a pre-incident baseline intended to stop attacks from spreading into physical harm. The coalition said private and local operators are not bound by binding operational directives, but issuing one would show what the government views as best practice.
Bipartisan senators Maria Cantwell and Mike Rounds urged White House AI czar Jay Clayton to develop a new national strategy addressing AI-assisted cyber threats. The senators said AI is rapidly growing the ability to discover and exploit software weaknesses, while defenses are not moving fast enough.
The senators said patches intended to protect infrastructure can require weeks or months for testing and installation, and autonomous AI agents can introduce new security risks. Jay Clayton was named to the AI role in early October and heads a task force of government officials charged with producing an AI policy by the end of January.
The senators pointed to an ongoing wave of cyber attacks in Japan hitting financial institutions and companies, causing major data leaks and breaches that look AI-driven. They also said rogue AI agents got into customers’ bank accounts, causing banks to worry that AI might start transactions without human involvement.
The senators asked Clayton’s task force to unite AI companies with federal agencies to secure digital infrastructure and create security standards for autonomous AI agents.
Secretary of State Marco Rubio said recent outages affecting the Federal Aviation Administration, United Airlines, and Microsoft Outlook are unconnected and show no evidence of nefarious actors. The White House did not immediately answer a request for comment on the senators’ letter.
Clayton was scheduled to appear before the House Committee on Homeland Security, with questions about AI’s role in recent outages expected on Wednesday.

Join the discussion