.svg){kind=logo}
TL;DR
AI can introduce security exposures that an existing ISO 27001 program has not yet assessed. The opportunity for leaders is to bring AI use into the information security management system, connect safeguards to evidence, and review them as systems change. ISO/IEC 27090:2026 offers complementary guidance on AI-specific threats; it does not replace ISO 27001.
A financial services company introduces an AI assistant to help employees find answers in internal documents. It starts as a productivity initiative. A few months later, the assistant can search sensitive files and is connected to a workflow that drafts customer responses.
Now imagine the board asks whether those new connections are covered by the organization’s security program. The company has an ISO 27001 certificate, but the team cannot immediately show who approved the assistant’s access or whether those permissions have been reviewed since the rollout.
That is not a reason to stop using AI. It is a reason to ask whether the information security management system (ISMS) has kept pace with how the business uses it. For CEOs of regulated enterprises, government suppliers, and critical infrastructure organizations, that answer can influence customer trust as much as operational resilience.
ISO 27001 and AI security: the question behind the certificate
ISO/IEC 27001:2022 provides the requirements for an ISMS. Its risk-based approach gives organizations a foundation for governing information security, but a certificate should not be assumed to cover every AI deployment. The scope of the certified ISMS matters, as do the risks the organization has assessed and the controls it has selected.
That distinction is becoming more important as AI moves from standalone tools into business processes. In its 2026 AI cybersecurity risk assessment, Ireland’s National Cyber Security Centre found that majority of public-sector respondents were planning wider AI use while raising concerns about issues including shadow AI and data leakage. Its survey was small—12 respondents—so the findings are a useful signal rather than a measure of the whole market.
For a CEO, the practical question is whether the organization can explain how a material AI use case is governed within its existing security processes. If the answer depends on an informal conversation or an outdated approval, the certificate alone will not resolve the uncertainty.
What ISO/IEC 27090 adds to the picture
Published in October 2026, ISO/IEC 27090:2026 provides guidance on security threats and compromises specific to AI systems. ISO describes it as complementary to ISO 27001 and ISO 27002, helping organizations understand where established security practices remain useful and where AI data, models, and workflows need additional attention.
The guidance covers threats such as data poisoning and model theft. It also recognizes that model updates and retraining can introduce new exposure if the data used has been corrupted. ISO 27090 is guidance, not a replacement for ISO 27001 or a guarantee of security.
The implication for leadership is straightforward: an existing security program is a starting point, but it needs an explicit view of the AI systems that matter to the business. That includes tools purchased from suppliers, not only models developed in-house.
Follow the access, not just the application
Going back to the AI assistant, the most consequential change may not be the tool itself, but the access it receives. An assistant that can read a policy library presents a different risk from one that can retrieve confidential client records or initiate actions in another system.
A practical review starts with the intended business use and the information the system can reach. Security and business owners can then assess whether access is proportionate, whether a human must approve consequential actions, and what should trigger a new review. A supplier’s change to data handling or an expansion of permissions may be enough to change the risk picture.
A question worth taking to the next executive review: If an AI assistant’s permissions changed tomorrow, who would know, who would assess the impact, and what evidence would show the decision was made?
Turn an AI security risk into an assurance trail
Consider a customer service assistant that can retrieve account information. The organization might identify a risk of inappropriate access to sensitive records and select an access control that restricts the assistant to approved data sources. The control owner can then arrange a periodic test of those permissions and retain the resulting access review as evidence.
If the test finds that the assistant can access a repository outside its approved scope, the finding should lead to an assigned action and a subsequent check. This example is illustrative; the right control and frequency depend on the system and the organization’s risk assessment.
Illustrative assurance trail
AI assistant access risk → approved access control → permission test → review evidence → action and retest
That trail matters because leadership can see the difference between a control that exists on paper and one that has been examined in practice. It also gives teams a way to revisit the decision as the assistant’s role changes, rather than waiting for the next certification audit.
Bring evidence together with 6clicks Intelligent GRC
6clicks Intelligent GRC helps organizations connect controls, tests, evidence, and actions within a broader governance program. In the assistant example, a team can use those relationships to review the access control alongside its supporting test and submitted evidence, making the assurance story easier to follow.
Hailey can automate control testing and evidence collection or assess submitted evidence against a test-specific validation guide and surface gaps for review. People remain accountable for interpreting the findings and reviewing the test result.
For organizations operating across multiple entities or sensitive environments, the advantage is a clearer view of assurance without asking leaders to reconstruct it from disconnected records. The next conversation can then focus on what needs attention, rather than where the evidence is.
Explore 6clicks Intelligent GRC to see how your team can connect the evidence behind its security controls.
Frequently asked questions
No. Coverage depends on the certified ISMS scope and how the organization assesses and manages relevant risks. Review each material AI use against that scope, especially when its access or business purpose changes.
No. ISO/IEC 27090:2026 offers guidance on AI-specific security threats and complements ISO 27001 and ISO 27002. ISO 27001 remains the information security management system requirements standard.
Depending on the risk and control design, useful evidence might include an approved access decision, a current permission review, and a record of any corrective action. The control owner should determine what is sufficient for the test and protect sensitive records appropriately.
6clicks Intelligent GRC can connect relevant controls with tests and supporting evidence. Hailey can help automate control testing and evidence collection or assess submitted evidence and identify gaps, while people review the output and make the final decision.
