AI & Cybersecurity
ISO/IEC 27090:2026 Published — What It Means for AI Security and EU Compliance
ISO/IEC 27090:2026 introduces international guidance for addressing AI-specific cybersecurity threats. It does not establish a new certification scheme or automatically confer conformity with the EU AI Act. CORVUS AI examines its legal significance, implications for NIS2, supplier due diligence, defence and dual-use technologies, and the practical steps organisations should take.
Complex law. Clear action.
Reviewed by Oleksandr Sobovyi, Founder & CEO of CORVUS AI — editorial responsibility statement below.
The new international standard addresses AI-specific cybersecurity threats without introducing a separate certification scheme or creating a presumption of conformity under the EU AI Act. Its practical significance lies elsewhere: it provides a reference point for assessing whether AI providers and deployers have adequately identified, evaluated and addressed security risks.
On 14 October 2026, ISO/IEC 27090:2026 was published under the title Cybersecurity — Artificial intelligence — Addressing security threats and compromises to artificial intelligence systems.
Developed by ISO/IEC JTC 1/SC 27, the standard provides guidance on security threats affecting AI systems throughout their lifecycle.
Its publication is relevant to AI developers, technology providers, organisations deploying third-party AI solutions, public authorities and operators of critical infrastructure.
The central question is not whether organisations can obtain another certificate. It is whether they can demonstrate that AI-specific cybersecurity risks have been identified, assessed and appropriately managed.
What ISO/IEC 27090 Does — and Does Not — Establish
ISO/IEC 27090 is a guidance standard rather than a standalone set of certifiable management-system requirements.
It addresses security threats and potential compromises specific to AI systems, including data poisoning and model theft, and complements established information security frameworks such as ISO/IEC 27001 and ISO/IEC 27002.
Two distinctions are particularly important.
First, ISO/IEC 27090 does not establish a separate accredited certification scheme.
Organisations should not confuse assessments against its recommendations with certification under ISO/IEC 27001 or ISO/IEC 42001.
Independent audits and contractual assessments may nevertheless use the standard as a technical reference.
Second, the absence of a separate certification scheme does not make the standard irrelevant.
Its recommendations may help organisations document the reasonableness and proportionality of their cybersecurity measures.
Relevant evidence may include AI asset inventories, threat models, risk assessments, security testing results, incident-response procedures and supplier assurance documentation.
Such evidence does not automatically establish legal compliance. Its significance depends on the applicable legal obligations, contractual commitments and circumstances of the system concerned.
Relationship with EU Law
1. EU AI Act: Supporting Cybersecurity Compliance
Article 15 of Regulation (EU) 2024/1689 establishes requirements concerning accuracy, robustness and cybersecurity for high-risk AI systems.
The Regulation expressly addresses threats involving training data poisoning, model poisoning, adversarial examples, model evasion, confidentiality attacks and model-related vulnerabilities.
ISO/IEC 27090 addresses a related area of cybersecurity risk and may assist organisations in developing and documenting appropriate protective measures.
However, compliance with ISO/IEC 27090 does not automatically establish compliance with Article 15 of the AI Act.
Under Article 40, a presumption of conformity may arise from harmonised standards, or relevant parts thereof, whose references have been published in the Official Journal of the European Union, within the scope covered by those standards.
ISO/IEC 27090 does not acquire that legal effect merely because it is an international standard.
European harmonisation work under the AI Act involves CEN-CENELEC JTC 21. International ISO standards and European harmonised standards must therefore be distinguished in legal and contractual documentation.
For high-risk AI systems, cybersecurity compliance must also be considered alongside other applicable obligations, including risk management, data governance, technical documentation, record-keeping and quality management.
The applicability and implementation deadlines of the relevant AI Act provisions must be assessed under the legislation in force at the relevant time.
Practical implication: ISO/IEC 27090 may support a cybersecurity compliance programme, but it cannot replace a legal assessment or the applicable conformity assessment procedure.
Official source: Regulation (EU) 2024/1689 — AI Act.
2. NIS2: AI Security Within Cybersecurity Risk Management
Article 21 of Directive (EU) 2022/2555 requires essential and important entities within its scope to implement appropriate and proportionate technical, operational and organisational cybersecurity risk-management measures.
These measures include supply-chain security, incident handling, business continuity and security in the acquisition, development and maintenance of network and information systems.
For organisations subject to NIS2 that integrate AI into important operational processes, ISO/IEC 27090 may provide an additional reference for identifying threats that conventional cybersecurity assessments do not sufficiently address.
However, the use of AI does not automatically bring an organisation within the scope of NIS2.
Applicability depends on the entity's activities, legal classification, relevant thresholds and national implementing legislation.
Practical implication: ISO/IEC 27090 may support risk-management documentation, but its application does not independently demonstrate compliance with Article 21 of NIS2.
Official source: Directive (EU) 2022/2555 — NIS2.
3. ISO/IEC 27001 and ISO/IEC 42001: Integrating Existing Management Systems
ISO/IEC 27001 provides a framework for information security management, while ISO/IEC 42001 establishes requirements for an AI management system.
ISO/IEC 27090 can complement these frameworks by helping organisations address cybersecurity threats arising from the specific characteristics of AI technologies.
Organisations with established management systems should consider incorporating relevant AI-specific risks into existing governance, risk assessment and control processes rather than automatically creating a parallel compliance structure.
Where Legal Expectations and Operational Practice May Diverge
Misconception 1: AI security standards concern only model developers.
Organisations deploying third-party AI solutions may also face cybersecurity, contractual and sector-specific obligations.
The allocation of responsibilities depends on the organisation's role, the applicable legal framework and the relevant contractual arrangements.
Outsourcing AI functionality does not necessarily eliminate the need to assess associated security risks.
Misconception 2: Traditional cybersecurity controls are sufficient.
Conventional controls remain essential, but AI systems introduce additional attack surfaces.
Training data manipulation, adversarial inputs and attempts to extract information from models may require dedicated threat modelling, testing and monitoring.
Misconception 3: A voluntary standard cannot have contractual consequences.
Although ISO/IEC 27090 is not generally binding merely by virtue of publication, its provisions may become relevant through procurement specifications, contractual obligations or references in applicable regulatory instruments.
Organisations should therefore distinguish voluntary adoption from obligations created by a particular legal or contractual framework.
Implications for Defence, Dual-Use Technologies and EU-Funded Projects
AI security is particularly relevant where systems operate in sensitive environments, process protected information or support critical functions.
Defence organisations, dual-use technology providers and participants in EU-funded projects may face extensive cybersecurity and supply-chain assurance requirements.
ISO/IEC 27090 may become an additional technical reference in these assessments.
However, there is no basis for assuming that its publication automatically introduces a new eligibility condition under the European Defence Fund, Ukraine Facility or defence procurement frameworks.
Any such requirement must be established by examining the applicable programme rules, procurement documentation, grant agreements or contractual provisions.
For suppliers, the strategic opportunity is to prepare documented evidence of AI security governance before specific customers request it.
What Organisations Should Do Now
1. Establish an AI asset inventory. Identify internally developed and third-party AI systems, models, datasets and relevant dependencies.
2. Document AI-specific threat models. Assess the relevance of data poisoning, model poisoning, adversarial inputs, confidentiality attacks and other threats based on each system's architecture and intended use.
3. Review supplier contracts. Evaluate provisions concerning security testing, incident notification, access to relevant documentation, vulnerability management and allocation of responsibility.
4. Assign accountability. Establish clear responsibility for AI cybersecurity risk management across technical, legal, procurement and governance functions.
5. Conduct a gap assessment. Compare existing security measures with the published ISO/IEC 27090:2026 recommendations and relevant legal requirements, recording identified gaps and remediation priorities.
The Strategic Conclusion
ISO/IEC 27090:2026 does not create a new EU legal obligation by itself.
Its significance is that it provides a specialised international reference for evaluating cybersecurity risks associated with AI systems.
As organisations increasingly rely on AI in sensitive operational environments, documented security governance may become more important in supplier assessments, procurement procedures, contractual negotiations and regulatory scrutiny.
The objective should not be to claim compliance with a voluntary standard as a substitute for legal compliance. It should be to establish a defensible, evidence-based approach to AI cybersecurity.
CORVUS AI — Legal and Regulatory Advisory
CORVUS AI supports organisations in assessing AI cybersecurity governance, regulatory obligations, supplier risks and contractual arrangements.
Our legal and regulatory gap assessments can examine existing AI security measures against the relevant provisions of the EU AI Act, NIS2 and ISO/IEC 27090:2026.
Disclaimer
This article has been prepared by CORVUS AI for general informational and educational purposes only. It is intended to make complex legal and regulatory developments easier to understand.
It does not constitute legal advice and does not create a professional adviser–client relationship. The information should not be relied upon as a substitute for advice based on the specific facts, circumstances and applicable law relevant to your organisation or project.
The article reflects our understanding of the law and regulatory framework as of the date of publication. Legislation, case law, regulatory guidance and administrative practice may subsequently change. While reasonable care has been taken in preparing this article, CORVUS AI does not warrant that the information is complete or remains current after the date of publication. We do not undertake to update this content.
To the fullest extent permitted by applicable law, CORVUS AI excludes liability for loss arising from reliance on this article. Nothing in this article constitutes an offer or solicitation to provide regulated legal services in any jurisdiction where doing so would be unlawful.
AI-assisted preparation: This article was prepared with the assistance of AI tools. Its legal analysis, conclusions and final text were subject to human review and editorial control and were reviewed and approved prior to publication by Oleksandr Sobovyi, Founder & CEO of CORVUS AI. CORVUS AI retains editorial responsibility for the published content.
For advice tailored to your organisation, project or specific circumstances, please contact CORVUS AI.
For enquiries: contact@corvusai.eu
