Top security agencies unveil joint recommendations for safe AI integration across critical systems

The National Security Agency (NSA) has collaborated with the Cybersecurity and Infrastructure Security Agency (CISA), the Australian Signals Directorate’s Australian Cyber Security Centre (ASD’s ACSC) and several international security bodies to release a new Cybersecurity Information Sheet (CSI) outlining how artificial intelligence (AI) should be safely incorporated into critical operational technology systems.

According to the joint statement, the rapid rise of AI offers major benefits in efficiency, decision-making and automation, but the same tools can also introduce new risks to essential infrastructure.

The agencies warned that poorly governed AI models could disrupt safety controls, compromise industrial processes, or expose sensitive operational data.

Titled "Principles for the Secure Integration of Artificial Intelligence in Operational Technology," the report urges operators of critical infrastructure to adopt a cautious and highly controlled approach.

The guidance outlines four core principles: fully understanding AI systems, evaluating whether AI is appropriate for an operational technology environment, establishing strict governance and assurance frameworks, and embedding security safeguards throughout AI-enabled systems.

The agencies emphasised key mitigations such as using separate AI systems for data processing, integrating human oversight for sensitive decisions, and ensuring comprehensive testing before deployment.

In its conclusion, the coalition of security agencies stressed that responsible AI integration will be essential to protecting national infrastructure while still allowing organisations to benefit from emerging technologies.