The Operational Technology Cybersecurity Coalition (OTCC) urges U.S. policymakers to make critical infrastructure resilience a central pillar of national AI security policy. In a new report, the coalition sets out 18 recommendations to help essential services withstand, contain and recover from AI-enabled cyberattacks. Developed through four summer roundtables involving nearly 70 representatives from critical infrastructure sectors, government, and the cybersecurity community, these recommendations focus on strengthening baseline cybersecurity, enabling AI-supported defenses in OT (operational technology) environments, and improving resilience during disruptions.
Titled ‘The Missing Pillar of AI Security: Critical Infrastructure Resilience,’ the report calls for closing regulatory gaps without duplicating existing requirements, directing funding to under-resourced entities, establishing an AI governance framework for OT environments and ensuring long-term access to AI-enabled cybersecurity tools. It also recommends funding the Cyber Response and Recovery Fund and reconstituting the Cyber Safety Review Board.
OTCC grounded its recommendations in three core principles covering prioritizing resilience, assuming compromise, and involving AI developers in OT defense. The coalition cited an August warning from CISA and federal partners that attackers are using AI-generated scripts to target industrial controllers in the energy, water, and manufacturing sectors, allowing adversaries to identify and exploit longstanding vulnerabilities faster and at greater scale.
“Congress and the Administration don’t need to start from scratch,” Michael Garcia, policy director of OTCC, said in a media statement. “Every recommendation in this report builds on authorities, programs, and institutions that already exist. What’s missing is a decision to treat operational resilience as part of AI security policy, and to fund it that way.”
While emerging AI capabilities introduce novel threats, operational technology (OT) defense still relies on operational resilience grounded in fundamental cybersecurity practices like access management, network segmentation, and zero trust. Because AI tools alone cannot compensate for missing baseline security, national AI security policy must measure success by whether critical infrastructure operators can sustain and rapidly restore essential services during an incident, a priority echoed in U.S. strategies like NSM-22 and the 2026 National Resilience Strategy, as well as recent U.K. guidance.
As AI accelerates vulnerability discovery, critical infrastructure operators must adopt an ‘assume compromise’ mindset rather than relying solely on patching. Roughly 70% of OT vulnerabilities reside in legacy or deeply embedded systems that cannot be patched or rebooted without disrupting continuous physical operations, and a shortage of specialized engineering talent creates severe bottlenecks in testing the over 12,000 ICS vulnerabilities tracked by CISA. Assuming compromise reframes OT defense around building resilience so infrastructure can withstand inevitable breaches.
Securing critical infrastructure in the AI era requires proactive collaboration among the federal government, AI labs, cybersecurity vendors, and OT owners and operators. AI companies must partner with OT and federal stakeholders to safely train models on sensitive infrastructure data, such as keeping protected electric utility data within owner environments or controlled test beds, and include OT partners in industry policy initiatives to encourage active defensive experimentation.
The OTCC report follows last week’s nationwide FAA ground stop and Microsoft outage, which caused nationwide chaos and briefly sparked fears of a coordinated cyberattack. “Although neither event was a cyberattack, both demonstrated the exact scale of disruption our adversaries are working to engineer, now with the added speed and sophistication of artificial intelligence. A software glitch is an inconvenience, but a deliberate, AI-driven strike on the operational technology powering our essential services would cause untold harm to Americans around the country and impact our national security.”
OTCC outlines 18 recommendations that build on existing authorities, programs, and institutions. By integrating AI into OT cybersecurity with appropriate safeguards, Congress and the Biden administration can take practical steps to protect the systems underpinning U.S. critical infrastructure.
To strengthen baseline cybersecurity against AI-enabled threats, the report calls for closing cross-sector regulatory gaps without duplicating requirements, establishing secure-by-design mandates and a cyber duty of care, funding under-resourced entities, enacting whistleblower protections, and introducing compensating controls for legacy, unpatchable systems.
For years, malicious cyber actors have compromised entities and OT systems through well-known security gaps. They exploited weak passwords, the absence of multi-factor authentication, overprivileged access, and inadequate network segmentation. AI will only exacerbate this challenge. Recent attacks show that bad actors, whether using AI-enabled tools or not, compromise systems through these same gaps. While cyber defenders could use AI to identify these flaws and address them using established practices, there is no guarantee they will. In other words, too few incentives or requirements exist to compel organizations to act, despite increased AI-enabled cyberattacks.
Meanwhile, organizations are simultaneously deploying their own AI agents and other non-human identities (NHIs) to fight back against AI-enabled attacks. Yet, they are adopting them at a rate that far exceeds their ability to govern them. In this new landscape, where NHIs can outnumber human employees by more than 100 to 1, the very definition of identity has changed. This creates a security concern in its own right by potentially having an exponential increase in insider threats within an organization. Therefore, any organizational cybersecurity framework must account for securing AI agents by extending common security controls such as zero trust and identity access management protocols.
The report identified that Congress and the Executive Branch must establish a national framework that aligns enforceable cybersecurity requirements, incentives, accountability, and financial assistance to increase the adoption of baseline cybersecurity and identity security controls across critical infrastructure. This includes mandatory, harmonized cybersecurity requirements across critical infrastructure; secure-by-design requirements for hardware and software producers to increase resilience at scale; liability protections for those that fall outside critical infrastructure sectors or hardware and software providers; and financial assistance to under-resourced entities.
The framework should establish a common cybersecurity floor based on CISA’s Cross-Sector Cybersecurity Performance Goals (CPGs), while allowing sector-specific regulators to impose additional, risk-based requirements. These requirements must also account for AI agents and other NHIs operating with privileged access to sensitive systems. This framework is not meant to add another layer of cybersecurity compliance, but to ensure every critical infrastructure sector meets a minimum level of security commensurate with the risks it poses to national security, public safety, and essential services.
Forging connective tissue between AI labs, the federal government, state and local governments, and OT partners will be critical to strengthening resilience against AI-enabled cyberattacks. AI developers possess unique knowledge of emerging model capabilities and potential misuse, while OT manufacturers, cybersecurity providers, and critical infrastructure operators understand the operational realities of securing industrial systems.
Yet, AI adoption rates within OT have been slow. Nozomi Networks found that fewer than 8% of organizations they surveyed have deployed AI across multiple OT cybersecurity functions. Congress, the Executive Branch, and private-sector entities should establish sustainable partnerships that bring these communities together to identify high-value applications of AI, test defensive capabilities in realistic OT environments, and accelerate the adoption of proven technologies across critical infrastructure.
These partnerships should include OT owners and operators, industrial automation providers, and cybersecurity service providers alongside AI developers, so that pilots address real-world deployment barriers, including legacy infrastructure, workforce constraints, interoperability, and the need to maintain operational continuity. Federal agencies should use pilot findings to develop practical, risk-based guidance that helps operators evaluate and procure effective AI-enabled defensive capabilities.
To enable AI-based defenses in OT environments, the coalition recommends establishing an OT-specific AI governance framework, scaling successful state and regional models, ensuring long-term access to AI-enabled capabilities, speeding the adoption of certified commercial off-the-shelf solutions, and testing autonomous AI deployments. It also urges renewed support for information sharing and analysis centers (ISACs), a public-private forum for AI developers and OT organizations, and stronger international coordination.
OTCC recognizes that malicious actors will increasingly use AI to identify vulnerabilities, compromise systems, and execute attacks with improved speed and scale. Even organizations that implement strong cybersecurity controls cannot eliminate every vulnerability. While implementing basic cybersecurity controls will reduce the overall risk landscape, it will not eliminate risk. Federal cybersecurity policy should emphasize the ability to withstand, contain, and recover from a successful attack.
OT owners and operators and the broader U.S. government should adopt zero trust principles for AI models and agents, embrace CISA’s CI Fortify approach, and assume compromise in either their system or in a system they rely upon. Congress and the Executive Branch should strengthen national cyber response and recovery capabilities, improve coordination between cybersecurity professionals and emergency responders, and support investments in operational resilience.
Federal resilience initiatives should encourage critical infrastructure owners and operators to regularly assess recovery readiness, including whether their security teams can detect incidents across operational assets, coordinate OT-specific response, and restore priority functions with minimal disruption. Policymakers should consider technical assistance and voluntary exercises that evaluate these capabilities in realistic operating conditions.
To enhance resilience during disruptions, the report advises funding the Cyber Response and Recovery Fund, reconstituting the Cyber Safety Review Board, improving incident management processes, and adopting Critical Infrastructure Fortify controls. In the coming months, OTCC plans to convene policymakers, AI developers, and the OT community to test and refine these proposals, while engaging Congress, federal, state, and local governments, and AI laboratories to drive implementation.
In conclusion, the OTCC report observed that the past several months have made clear that AI-enabled threats to OT environments are no longer hypothetical. Attackers are leveraging AI to discover vulnerabilities, compromise devices, and target industrial controllers at a speed and scale that traditional defenses were not designed to meet.
The recommendations rest on a simple premise that operational resilience, grounded in fundamental cybersecurity practices and enhanced with AI-enabled defenses, forms the foundation of OT defense in this new era. Many of these actions build on existing authorities, programs, and institutions, requiring prioritization, sustained funding, and the will to act before the next crisis strikes. While CISA is responsible for implementing several of these measures, the agency faces low staffing levels and insufficient resources. Rebuilding and supporting CISA and the SRMAs is imperative for national cybersecurity, because as AI continues to lower the barrier to sophisticated attacks, the cost of delay will only grow.
The coalition is firm in its commitment to advancing these recommendations beyond this report. In the coming months, the OTCC will publish follow-up papers examining pressing issues in greater depth, convene policymakers, AI developers, and the OT community to test and refine these proposals, and engage directly with the Executive Branch, Congress, state and local governments, and AI labs to drive implementation. While the window to build resilience before the next major AI-enabled cyberattack is narrowing, it remains open, and the OTCC looks forward to working with public and private sector partners to seize this opportunity.
Earlier this month, the OTCC, along with 34 critical infrastructure organizations, urged U.S. congressional leaders to include a ten-year reauthorization of the Cybersecurity Information Sharing Act of 2015 (CISA 2015) in the final Fiscal Year 2027 National Defense Authorization Act (NDAA). The law enables voluntary sharing of cyber threat indicators and defensive measures between private-sector organizations and the federal government, providing liability protections and legal certainty for participating organizations.


