Artificial intelligence is rapidly becoming part of normal business operations.
Employees are using generative AI to draft documents, analyse information, write software, summarise meetings and support decision-making. AI functionality is also increasingly being integrated into the software, cloud platforms and business applications that organisations already use.
For organisations that already hold ISO 27001 certification, as well as those currently implementing an Information Security Management System (ISMS), this presents an important question:
Has your ISMS been designed or enhanced to address the information security risks associated with the use of AI?
The good news is that organisations do not necessarily need to create a completely separate information security framework simply because they use AI.
ISO 27001 already provides a structured approach for understanding organisational risks, protecting information, managing suppliers and continually improving security controls.
However, AI can introduce new risks and information flows that may not have been considered when an ISMS was first implemented.
For organisations already certified to ISO 27001, the challenge is ensuring that the ISMS continues to evolve as AI is introduced. For organisations implementing ISO 27001, there is an opportunity to consider AI risks from the beginning and build appropriate controls into the ISMS rather than trying to address them later.
This article provides practical guidance for businesses at both stages and explains what ISO 27001 auditors are likely to look for when assessing how AI-related information security risks are managed.
The first question should be:
How is AI being used within our organisation, and what risks does that create for our information?
This applies whether your organisation:
For organisations implementing ISO 27001, AI should be considered when establishing the context, scope and risks of the ISMS.
For organisations already certified, AI should be considered as part of the ongoing review and improvement of the ISMS.
The objective is the same:
Ensure the organisation understands how AI affects the confidentiality, integrity and availability of information and has implemented controls appropriate to the associated risks.
AI security guidance is increasingly being developed to complement established information security practices rather than replace them. ISO/IEC 27090, currently under publication, specifically focuses on threats and compromises affecting AI systems and is intended to complement established practices such as ISO/IEC 27001 and ISO/IEC 27002.
Before risks can be managed, the organisation needs to understand how AI is actually being used.
This should be a consideration for both:
Organisations Implementing ISO 27001
AI use should be identified as part of understanding:
Organisations Already Certified
Existing AI use should be reviewed to determine whether changes to the organisation's technology, processes or information flows have created new risks that need to be addressed through the ISMS.
Employees may already be using AI tools for:
AI functionality may also exist within software that the organisation has already approved.
Examples may include:
The organisation may not consider these systems to be AI systems because they were originally implemented for another purpose.
However, if new AI functionality changes how information is processed, stored or transferred, the associated risks should be considered.
An auditor may ask:
The organisation does not necessarily need a complicated register of every AI feature.
However, it should be able to demonstrate that it understands where AI presents a material information security risk.
For many organisations, an AI register or an extension of the existing asset register may be a practical solution.
Australian Government guidance also identifies AI registers, clear accountability and supply chain accountability as important governance practices and recommends keeping records that support audits, reviews and ongoing improvement.
For organisations implementing ISO 27001, AI risks should be considered as part of the initial information security risk assessment.
For organisations already certified, the introduction or increased use of AI should trigger a review of whether new risks have been created.
The organisation does not necessarily need a completely separate AI risk management process.
Instead, the existing ISO 27001 risk assessment methodology can be used.
AI may introduce risks associated with:
Risk
An employee uploads confidential customer information into an unapproved public AI platform.
Potential Consequences
Existing Controls
Additional Treatment
This approach demonstrates that AI is being managed through the organisation's existing ISMS rather than as a separate issue.
Auditors are unlikely to expect every organisation to identify the same AI risks.
Instead, they will generally be interested in whether:
For organisations implementing ISO 27001, auditors will look for AI risks to be incorporated into the ISMS where relevant.
For certified organisations, they may examine whether the risk assessment has evolved to reflect changes in how AI is used.
ISO 27001 requires organisations to understand the internal and external issues that are relevant to the purpose of the organisation and affect its ability to achieve the intended outcomes of the ISMS.
For businesses implementing ISO 27001, AI may already be a relevant internal or external issue.
For example:
Internal Issues
External Issues
For certified organisations, management should consider whether the growing use of AI represents a change to the organisation's internal or external environment.
One of the first questions organisations implementing an ISMS should consider is:
Who is responsible for managing the risks associated with AI?
The answer will depend on the organisation.
Responsibility may involve:
However, responsibility should not be unclear.
For a small organisation, responsibility may sit with one person or a small management team.
For a larger organisation, responsibilities may be distributed across multiple functions.
The important point is that the organisation understands:
Australian Government guidance similarly recommends clearly assigning accountability for AI systems, including responsibilities for third-party systems, testing, oversight and ongoing improvement.
The use of AI can create significant risks where Personally Identifiable Information (PII) or personal information is involved.
This should be considered during ISO 27001 implementation and reviewed regularly by certified organisations.
Businesses should understand whether employees are permitted to upload:
Where personal information is processed through AI, organisations should consider:
Can Personal Information Be Uploaded?
If yes, under what circumstances?
Is Information Retained?
How long does the supplier retain prompts, documents or outputs?
Is Information Used for Training?
Can customer information contribute to model training or service improvement?
Where Is Information Processed?
Are there data residency or international transfer considerations?
Can Information Be Deleted?
Does the organisation have the ability to remove information where required?
Who Can Access the Information?
Can supplier employees or sub-processors access it?
Auditors may seek evidence that the organisation has considered AI within its privacy and PII protection arrangements.
This may include:
Privacy is not the only concern.
For many businesses, one of the greatest risks associated with AI is the potential exposure of intellectual property.
Employees may use AI to:
This creates an important question:
What happens to the information after it is uploaded into the AI platform?
Businesses should consider controls relating to:
An important issue for organisations is that they may possess information without owning it.
For example, a customer may provide:
The organisation may have access to this information to provide a service.
However, that does not automatically mean the organisation has the right to upload the information into an external AI platform.
This is where information security, legal and contractual requirements intersect.
An auditor may ask:
"Are employees permitted to upload source code into AI tools?"
"Can customer information be entered into generative AI?"
"How do employees know what information they can and cannot upload?"
The organisation should be able to explain:
AI suppliers should be considered within the organisation's supplier management processes.
This applies to:
New AI Suppliers
A business may subscribe to a new AI platform.
Existing Suppliers Introducing AI
An existing software provider may introduce new AI functionality.
Both scenarios can change the organisation's information security risk profile.
For organisations implementing ISO 27001, AI suppliers should be identified during supplier and information flow analysis.
For organisations already certified, changes to supplier services should trigger consideration of whether additional risk assessment or supplier review is required.
Information Handling
Model Training
Security
Sub processors
Incident Management
Auditors may review whether:
A common risk arises where employees independently sign up to AI services using corporate email addresses or credit cards.
This can bypass established supplier management processes entirely.
Businesses implementing ISO 27001 should consider AI when designing their information classification scheme.
Businesses already certified should consider whether their existing classification scheme provides sufficient guidance for AI use.
Employees may understand that information is classified as:
However, they may not understand what those classifications mean when using AI.
Organisations should consider providing practical guidance with exact requirements t0 reflect the organisation's own risk assessment.
The objective is to provide employees with practical guidance.
Policies and employee awareness are important.
However, businesses should not rely solely on telling employees:
"Do not upload confidential information."
Depending on the risks, organisations should also consider technical controls.
These may include:
The Australian Signals Directorate's guidance for AI application development includes controls addressing data retention and use, preventing organisational data from being used for AI training without appropriate consent, and content filtering to detect and block sensitive data exposure.
A small business using AI to generate public marketing content may require very different controls to an organisation using AI to:
The organisation's risk assessment should determine the appropriate level of control.
Where an organisation operates, develops or integrates AI systems, additional technical controls may be required.
This is particularly important where AI can access:
The principle of least privilege should continue to apply.
Businesses should consider:
A particular risk arises when an AI assistant has broad access to organisational information.
An employee may not normally have access to a document.
However, if an AI system can access the document and does not correctly enforce permissions, the employee may be able to obtain the information indirectly.
Organisations should therefore consider whether AI integrations preserve existing access control requirements.
Businesses implementing ISO 27001 should consider AI-specific threats during risk assessment where AI is used.
Certified organisations should review whether existing risk assessments adequately address these threats.
Examples include:
Prompt Injection
Attempts to manipulate an AI system into ignoring intended instructions or security controls.
Excessive Permissions
AI systems receiving unnecessary access to organisational information or systems.
Data Leakage
Sensitive information being disclosed through AI prompts or outputs.
Model Poisoning
Manipulation of training data or models that affects the behaviour or integrity of an AI system.
Insecure Integrations
AI systems connected to internal systems without appropriate security controls.
Unauthorised Actions
AI agents performing actions without sufficient approval or human oversight.
AI-Generated Errors
Employees relying on inaccurate AI-generated information without appropriate review.
Not every organisation will face all of these risks.
The objective is to determine which risks are relevant to the organisation's actual use of AI.
ISO/IEC 27090 highlights AI-specific threats such as risks affecting training data and models, while emphasising that established cybersecurity practices remain essential.
Organisations implementing an ISMS should consider AI when developing relevant policies and procedures.
Certified organisations should review whether existing policies remain adequate.
This does not necessarily mean creating a lengthy standalone AI policy.
Depending on the organisation, AI requirements may be incorporated into:
Some organisations may also benefit from a dedicated AI Acceptable Use Guideline.
Employees should understand:
Training should include realistic examples. For example:
Can an employee upload a customer complaint into AI for summarisation?
Can a software developer upload proprietary source code to an AI coding assistant?
Can a salesperson use AI to improve a customer proposal?
Can an employee upload a spreadsheet containing customer contact information?
These examples are often more effective than generic instructions.
Employees often see AI as a productivity tool rather than an information security issue.
This makes targeted awareness particularly important.
Auditors may interview employees and ask:
"Can you use AI to help with your work?"
"Which AI tools are you allowed to use?"
"Can you upload customer information?"
"Can you upload confidential documents?"
"What should you do if you are unsure?"
If employees provide inconsistent answers, the organisation may have difficulty demonstrating that controls are operating effectively.
Businesses implementing ISO 27001 should ensure their incident management processes are capable of managing AI-related incidents.
Certified organisations should review whether their existing process remains suitable.
Examples may include:
The organisation should be able to answer:
"What happens if an employee accidentally uploads confidential information into an AI platform?"
A response may involve:
For organisations implementing ISO 27001, AI should be considered when determining applicable controls and developing the Statement of Applicability.
For certified organisations, the introduction of AI may require a review of whether existing control decisions remain appropriate.
The introduction of AI does not automatically mean additional controls must be selected.
However, organisations should consider whether AI-related risks affect the applicability or implementation of existing controls.
Particular areas may include:
An auditor may review whether the organisation's Statement of Applicability remains consistent with the way AI is actually being used.
For organisations implementing ISO 27001, internal audits should eventually examine whether AI-related controls are operating effectively.
For organisations already certified, AI should be included in the internal audit programme where it represents a relevant risk.
Internal audit questions may include:
Governance
Risk Management
Information Protection
Supplier Management
Technical Security
Awareness
Incident Management
This allows organisations to identify weaknesses before a certification or surveillance audit.
Whether an organisation is preparing for its first certification audit or maintaining certification, an auditor will generally be interested in whether AI-related information security risks have been appropriately managed.
Auditors will not necessarily expect every organisation to have:
The level of control should be appropriate to the organisation and its risks.
However, auditors are increasingly likely to expect evidence that the organisation understands how AI affects its information security environment.
Context
Has AI been considered as an internal or external issue where relevant?
Risk Assessment
Have AI-related information security risks been identified and assessed?
Information Classification
Do employees understand what information may be provided to AI systems?
PII and Privacy
Has the organisation considered how personal information is processed?
Intellectual Property
Are confidential information and proprietary assets protected?
Suppliers
Have AI providers been identified and assessed?
Legal and Contractual Requirements
Has the organisation determined whether information can legally and contractually be provided to AI services?
Technical Controls
Are AI systems and integrations appropriately secured?
Awareness
Do employees understand acceptable AI use?
Incident Management
Can the organisation identify and respond to AI-related incidents?
Monitoring
Are controls and AI suppliers reviewed?
Improvement
Does the organisation learn from incidents, changes and emerging risks?
Whether you are starting your ISO 27001 journey or preparing for your next surveillance audit, consider the following questions.
Understand Your AI Use
☐ Do we know where AI is being used?
☐ Have we considered AI functionality within existing software?
☐ Do we understand what information is being provided to AI systems?
Governance
☐ Is responsibility for AI clearly defined?
☐ Do we have a process for approving AI systems?
☐ Have we considered AI within our ISMS governance processes?
Risk Management
☐ Have AI-related information security risks been assessed?
☐ Have appropriate risks been added to the risk register?
☐ Have risk treatment decisions been documented?
Information Protection
☐ Do employees understand what information can be uploaded?
☐ Are PII and personal information protected?
☐ Is intellectual property protected?
☐ Are confidential customer assets protected?
Suppliers
☐ Have AI suppliers been assessed?
☐ Do we understand where information is processed?
☐ Do we understand whether information is retained?
☐ Do we know whether information is used for model training?
Legal and Contractual Requirements
☐ Are we legally permitted to provide information to AI suppliers?
☐ Do customer contracts impose restrictions?
☐ Have privacy obligations been considered?
Technical Security
☐ Are AI integrations appropriately secured?
☐ Is access restricted according to least privilege?
☐ Are technical controls required to prevent data leakage?
Awareness
☐ Have employees received practical AI security guidance?
☐ Would employees know what to do if they were unsure?
Incident Management
☐ Can our incident management process manage AI-related incidents?
Audit and Review
☐ Has AI been considered within our internal audit programme?
☐ Have relevant policies and procedures been reviewed?
☐ Does our Statement of Applicability remain appropriate?
For organisations implementing ISO 27001, the goal should be to build an ISMS that reflects how the organisation operates today — including its use of AI.
For organisations already certified, the goal should be to ensure the ISMS continues to evolve as the organisation changes.
The purpose of an ISMS is not to create a static collection of policies and procedures.
It is to provide a structured framework for identifying and managing changing information security risks.
As organisations:
the ISMS should evolve accordingly.
A practical approach for both new and certified organisations is:
Step 1: Identify AI Use
Understand where AI is currently being used across the business.
Step 2: Assess the Risks
Apply your information security risk assessment methodology.
Step 3: Review Information Flows
Understand what information is being provided to AI systems and what information they can access.
Step 4: Review Suppliers
Assess AI providers and AI functionality within existing suppliers.
Step 5: Review Legal and Contractual Requirements
Confirm what information can legally and contractually be provided to AI systems.
Step 6: Update or Implement Controls
Develop policies, procedures and technical controls appropriate to the risks.
Step 7: Train Employees
Provide practical guidance based on real business scenarios.
Step 8: Review Effectiveness
Include AI within monitoring, internal audits and management review where relevant.
Artificial intelligence is changing how organisations create, process and share information.
Whether your organisation already holds ISO 27001 certification or is beginning the journey towards implementation, AI should be considered as part of your Information Security Management System.
For organisations implementing ISO 27001, this is an opportunity to build AI considerations into the ISMS from the beginning.
For certified organisations, it is an opportunity to review whether the ISMS has kept pace with changes in technology and business operations.
The key question is not:
"Do we need a separate management system just because we use AI?"
The more useful question is:
"Have we designed or enhanced our ISMS to manage the risks created by the way we use AI?"
Organisations should consider AI within:
Auditors will increasingly look beyond whether an organisation has an AI policy.
They will want to see evidence that the organisation understands how AI is actually being used and that its ISMS appropriately manages the associated information security risks.
Whether you are implementing ISO 27001 for the first time or maintaining an established ISMS, AI should not sit outside your information security management processes. If it processes, accesses or influences your organisation's information, it should be considered within your ISMS.
By addressing AI as part of ISO 27001 implementation and continual improvement, businesses can build a stronger foundation for certification while continuing to take advantage of the significant opportunities that AI can provide.
DQS Product Manager - Cyber Security and auditor for the ISO 9001, ISO 27001 standards and information security management systems (ISMS) with extensive experience in software development.
