CISA Publishes Security Guidance for Using AI in OT

A consortium of global agencies has published guidance on how to defend AI deployments in operational technology, emphasising that AI and OT are two of the most sensitive attack surfaces. The guidance, a 25-page document, notes that large language model deployments can boost efficiency and decision-making but also bring significant risks such as OT process models drifting over time or safety-process bypasses.

It aims to help OT operators understand AI, establish governance and assurance frameworks, and embed safety and security practices into AI integrations, with AI used to analyse critical data, detect anomalies in systems like SCADA, and optimise workflows.

The guidance warns that AI in OT must be bounded by understanding of cause-and-effect, risk tolerances, and the priority of safety and uptime, and it recommends education, careful consideration of the OT business case, data minimisation, and the incorporation of human-in-the-loop oversight to ensure failures do not disrupt critical operations.

According to The guidance was authored by the US's CISA, FBI, and NSA Artificial Intelligence Security Center; the Australian Signals Directorate's Australian Cyber Security Centre; the Canadian Centre for Cyber Security; the German Federal Office for Information Security; the Netherlands National Cyber Security Centre; the New Zealand National Cyber Security Centre; and the UK's National Cyber Security Centre.