
Cybersecurity — Artificial intelligence (AI) — Addressing security threats and compromises to AI systems - guidance for organisations seeking to protect AI systems against security threats across their life cycle
AI systems face security threats that require protection beyond conventional cyber security measures. From manipulated training data to attempts to extract AI models through seemingly legitimate queries, these threats can be difficult to detect and may undermine a system’s security and reliability.
ISO/IEC 27090 helps organisations understand these risks, assess their potential impact and consider appropriate detection and mitigation measures. As AI adoption grows, this guidance supports informed security decisions throughout development, deployment and ongoing operation, including model updates and retraining.
For organisations developing or using AI, ISO/IEC 27090:2026 offers a reference for reviewing AI security risks and considering appropriate detection and mitigation. It complements ISO/IEC 27001 and ISO/IEC 27002 by focusing on the additional protection that AI data, models and workflows may need.
If you are interested in purchasing a multi-user licence for this standard, please get in touch with us to discuss your requirements. You will receive a watermarked PDF which you will be able to store on an internal network, with access restricted to the predefined number of simultaneous users. Purchasing a multiuser licence grants you permission to print copies of the standard, up to the specified number of users.