One-sentence summary
ISO/IEC 27090:2026 is an international standard published in October 2026 on cybersecurity — AI — addressing security threats and compromises to AI systems, providing requirements for responses when AI systems are attacked, breached or tampered with.
Standard Overview
| Item | Detail |
|---|---|
| Standard number | ISO/IEC 27090:2026 |
| Title | Cybersecurity — Artificial intelligence (AI) — Addressing security threats and compromises to AI systems |
| Body | ISO/IEC |
| Status/Date | Published, October 2026 (Edition 1) |
| ICS | 35.030 (IT security) |
| Price | CHF 204 |
Scope
- For organizations building, deploying and operating AI systems, covering security threats and post-compromise responses across the AI lifecycle.
- Applies to enterprises and service providers embedding AI security into cybersecurity management.
- Complements ISO/IEC FDIS 27091 (AI privacy protection) across the security and privacy lines.
Key Clauses
| Clause area | Content | Affected party |
|---|---|---|
| Threat identification | Identify security threats and attack surfaces of AI systems | AI system builders |
| Intrusion response | How to respond and recover after breach or anomaly | AI operators |
| Tampering protection | Prevent model and data tampering | Model providers |
| Security governance | Embed AI security into organizational cybersecurity processes | Enterprise security teams |
Compliance Advice
- Use 27090 as the baseline to map AI system threat lists and incident response.
- Cross-check with 27091 (privacy protection) to meet security and privacy requirements separately.
- Cite the version and publication date (2026-10, Edition 1).
Official Sources
- ISO: ISO/IEC 27090:2026