# AI Usage Policy

> **Document control**  
> Owner: [POLICY_OWNER_ROLE, e.g. AI Officer]  
> Approved by: [APPROVER_NAME_AND_ROLE]  
> Version: [VERSION]  
> Effective date: [EFFECTIVE_DATE]  
> Next review: [NEXT_REVIEW_DATE]

> **AI Compliance Starter Kit.** This policy turns the AI policy into day-to-day work: approved services, handling of inputs and outputs, transparency under Art. 50, human oversight, reporting channels and the deployer obligations under Art. 26 of Regulation (EU) 2024/1689. It is addressed to all employees. Sections that apply only to certain roles carry a note in italics and are omitted where the role does not apply.

## 1. Legal and Normative Basis

ISO/IEC 42001:2023, Annex A control A.9.2 (responsible use of AI systems).

Regulation (EU) 2024/1689 (AI Regulation), in particular:

- Art. 4, AI literacy of staff who use AI systems.
- Art. 5, prohibited AI practices.
- Art. 50, transparency obligations, including the disclosure of artificial or manipulated content and of interaction with an AI system.
- Art. 26 in conjunction with Art. 14, obligations of deployers and human oversight of high-risk AI systems.

The handling of personal data when using AI is additionally governed by data protection law (GDPR), and the handling of information by the Information Security Policy. This policy carries the AI Policy into day-to-day practice and derives its authority from it.

## 2. Purpose and Scope

This AI Usage Policy sets out the rules for the use of artificial intelligence in the daily work of **[YOUR_ORGANISATION_NAME]**. It turns the principles of the AI Policy into concrete rules and draws a clear line between what is permitted and what is not, so that AI supports the work of the organisation without putting its information, its people or its legal standing at risk.

This policy applies to:

- All members of staff, whether permanent, temporary or part-time.
- External personnel, such as contractors, consultants and temporary staff, who use AI on behalf of the organisation.
- Every AI system and service used for work, including standalone AI assistants (for example ChatGPT, Microsoft Copilot, Google Gemini or Claude), AI features built into other software, and AI image, audio or code tools.

Staff familiarise themselves with this policy, acknowledge it and comply with its provisions. Where a person acts as the human oversight of a high-risk AI system, the specific arrangements recorded for that system apply in addition to this policy. A violation of this policy is handled through the disciplinary process of **[YOUR_ORGANISATION_NAME]**.

## 3. Approved AI Systems

Staff use only AI systems and services that **[YOUR_ORGANISATION_NAME]** has approved for the purpose at hand. Approval makes sure that a service is checked against the prohibited practices, the data protection requirements and the security requirements before it is relied upon for work.

### 3.1 Approved Services Only

- The approved AI services, together with the conditions of their use and the classification of data they may process, are kept in the catalogue of approved IT and AI services. Staff check the catalogue before using an AI service for a new type of task.
- Where an approved service is available in both a public version and a protected enterprise version, the enterprise version is used for work (for example an enterprise AI assistant with a data processing agreement rather than a free public tool).
- AI features built into software already in use are treated like any other AI service: they are used within their approved purpose, and features that send data to an external provider are only enabled where that is approved.

### 3.2 Requesting a New AI Service

A new AI service is taken into use only after it has been requested and approved through the established approval process. Staff who see a need for a tool that is not yet approved raise a request rather than using the tool on their own initiative, so that an approved alternative can be evaluated and, where suitable, added to the catalogue.

### 3.3 Shadow AI

The use of an AI service outside this approval, for example creating a private account on a public AI tool to handle work data, is not permitted. Any such use that comes to light is reported to the person responsible for AI, who assesses it and either brings it into the AI inventory under approved conditions or has it discontinued. New or changed AI applications in a work area are reported so that the AI inventory stays complete.

## 4. Rules of Use

### 4.1 Permitted Use

AI is used within the purpose for which it is approved and as a support to the work of the person using it, for example to draft and summarise text, to translate, to brainstorm ideas, to explain a topic or to assist with code. The person using an AI system remains responsible for the result and for any decision based on it. The output of an AI system is treated as a proposal that is reviewed and adapted; it is not adopted unchecked as a final result.

### 4.2 Prohibited Use

The following uses of AI are not permitted:

- **Prohibited practices:** Any use that falls under the practices prohibited by Art. 5 of the AI Regulation, such as manipulative or exploitative techniques that materially distort a person's behaviour, social scoring, the untargeted scraping of facial images, generating or manipulating intimate depictions of identifiable persons without their explicit consent and generating child sexual abuse material (both from 2 December 2026), risk assessment of individuals to predict criminal offences based solely on profiling or personality traits, real-time remote biometric identification in publicly accessible spaces for the purposes of law enforcement, or emotion recognition and biometric categorisation where the regulation forbids them.
- **Confidential and personal data in unapproved tools:** Entering confidential, classified or personal data into an AI service that has not been approved for that data, in particular free public AI tools. This also covers third-party trade secrets, source code and documents from the AI supply chain, for example from providers or conformity assessments.
- **Unchecked decisions about people:** Using the output of an AI system as the sole basis for a decision that has a legal effect on a person or affects them similarly (for example in recruitment, performance assessment or the handling of a customer request), without a review by a competent person.
- **Deception:** Passing off AI-generated content as human where this misleads the recipient, or creating content that impersonates a real person or their voice without a legal basis.
- **Circumventing controls:** Using AI to bypass security or compliance controls, or feeding another person's credentials, security information or protected material into an AI service.

### 4.3 Handling of Inputs and Data

Before data is entered into an AI service, its sensitivity is checked. The following rules apply:

- Confidential, classified or personal data is entered only into an AI service that has been approved for that classification. Free public AI tools are treated as if their inputs may become known to others.
- For sensitive work, the approved enterprise instances are used, because they keep the data within the organisation's agreement and do not use it to train the provider's models.
- Trade secrets, source code, unpublished business figures and the personal data of customers or colleagues are not entered into an AI service unless that service is expressly approved for such data.
- The rules of the Information Security Policy and the data protection documentation of **[YOUR_ORGANISATION_NAME]** apply in full to every input into an AI service.

### 4.4 Verification of Output

The output of an AI system is checked before it is used, because AI systems can produce results that are fluent and convincing yet incorrect. The following apply:

- **Accuracy:** Facts, figures, quotations, legal references and sources produced by AI are verified against a reliable source before they are relied upon.
- **Bias and fairness:** Output that concerns or affects people is checked for unjustified bias or discriminatory assumptions.
- **Rights in the material:** AI output is checked so that it does not reproduce third-party material in a way that infringes copyright or other rights.
- **Documentation:** Where the output is used for an external or consequential purpose, the review is documented so that it is traceable who checked the result and on what basis.

## 5. Transparency and Disclosure (Art. 50)

**[YOUR_ORGANISATION_NAME]** makes the use of AI transparent where the AI Regulation so requires. The following disclosure duties apply:

- **Artificial or manipulated content (deepfakes):** Image, audio or video content that is generated or manipulated by AI and appears deceptively real is disclosed as artificially generated or manipulated, for example a marketing image created with an AI image tool or a synthetic voice-over.
- **Interaction with an AI system:** Where people interact with an AI system such as a customer-facing chatbot, they are informed that they are dealing with an AI system, unless this is obvious from the circumstances.
- **AI-generated text on matters of public interest:** Text generated by AI and published to inform the public on matters of public interest is disclosed as artificially generated, unless it has undergone human review and a person or the organisation holds editorial responsibility for it.

### 5.1 Approval of Published Content

AI-generated content that is intended for external publication (such as a press release, a website text, a customer communication or a social-media post) is reviewed and approved before it is released. The review confirms the accuracy of the content, the disclosure required above and the rights in any material used. The disclosure is placed so that it is noticeable to the recipient (for example a clearly readable label on a synthetic image).

### 5.2 Timing and Accessibility of the Information

Where information is owed under Article 50(1) to (4), it is provided at the latest at the time of the first interaction or exposure, in a clear and distinguishable manner. It is reachable through assistive technology, understandable on its own and recognisable without relying on colour; for digital interfaces EN 301 549 applies.

For evidently artistic, creative, satirical, fictional or analogous works, the disclosure is limited to revealing the presence of generated or manipulated content in an appropriate manner that does not hamper the display or enjoyment of the work.

## 6. Human Oversight

Where the use of an AI system has a consequential effect on people or on the organisation, a competent person reviews the output before it is acted upon, so that an AI result is never put into effect automatically without the chance for a person to check and, if necessary, correct it.

For high-risk AI systems, the human oversight assigned to the system applies (AI Regulation, Art. 26 in conjunction with Art. 14): the person responsible for oversight understands the capabilities and limits of the system, monitors its operation, recognises signs of malfunction or unexpected output, is able to disregard or override an output and can stop the system where necessary. The arrangements for oversight are recorded with the classification of each high-risk system.

## 7. Reporting Incidents and Concerns

Staff report the following through the AI incident reporting channel of **[YOUR_ORGANISATION_NAME]**, promptly and without waiting for certainty:

- Malfunctions of an AI system, and output that is incorrect, harmful or discriminatory.
- Any suspected prohibited use, and any case where confidential or personal data may have been entered into an unapproved AI service.
- Signs that an AI system is being used outside its approved purpose or is producing results that affect people unfairly.

Reports made in good faith are handled without disadvantage to the person reporting. The person responsible for incident reporting assesses each report, initiates the necessary corrective action, and forwards serious incidents to the competent authorities where the AI Regulation requires it.

Concerns about the use of an AI system at **[YOUR_ORGANISATION_NAME]** can be raised through the AI reporting channel ([AI_REPORTING_CHANNEL, e.g. reporting form or mailbox]), either confidentially or anonymously. The route is also open to contracted persons, and a report made in good faith leads to no disadvantage (ISO/IEC 42001 A.3.3).

## 8. Right to Explanation (Art. 86)

Where a released high-risk AI system makes a decision that produces legal effects concerning a person or similarly significantly affects them, that person has the right to obtain from **[YOUR_ORGANISATION_NAME]** a clear and meaningful explanation: the role the system played in the decision-making procedure and the main elements of the decision.

Explanation requests are received through the channel set up for them ([EXPLANATION_CHANNEL, e.g. website form or mailbox]) and are handled separately from the complaint channel under Article 85. **[YOUR_ORGANISATION_NAME]** answers each request within 30 days using a template that states the system's role, the input data used, and the responsible unit, and records the answer for each request.

## 9. Operation and Support

Each approved AI system has a defined support path through which staff report faults and anomalies. The path, the responsible point of contact and the promised response time are recorded per system in the AI register; for purchased systems, the provider's commitments are held there.

Repairs and updates follow a defined process. Before an update is adopted, the responsible point of contact checks whether the system's behaviour changes in a safety-relevant way. Where an update changes a system's behaviour in a way material to its use, the affected users are informed before or upon adoption.

## 10. Deployer Obligations for High-Risk Systems (Art. 26)

*Only where the organisation deploys high-risk AI systems; otherwise this section is omitted.*

As the deployer of high-risk AI systems, **[YOUR_ORGANISATION_NAME]** carries out the following obligations under Article 26 of the AI Regulation. They are handled by the responsible roles and complement the general rules above.

Each high-risk system is used only in accordance with the provider’s instructions for use; permitted use cases, limits and responsibilities are defined internally. Human oversight is assigned to a person with the necessary competence, training and authority, who receives the support needed to intervene in or stop the system.

To the extent **[YOUR_ORGANISATION_NAME]** controls the input data of a high-risk system, it ensures the data is relevant and sufficiently representative for the system’s intended purpose.

The operation of each high-risk system is monitored on the basis of the instructions for use, and the provider is informed where relevant. Where there is reason to believe that use may present a risk to health, safety or fundamental rights, use is suspended and the provider or distributor and the market surveillance authority are informed without delay; serious incidents are reported.

For each high-risk system it is defined which events are logged, who may access the logs and where they are stored. Logs under the organisation’s control are kept for at least six months, unless a longer period applies; once the retention period ends they are deleted, and where data protection law requires earlier deletion of personal data, that obligation takes precedence.

Before a high-risk system is put into service at the workplace and the organisation’s own workers become subject to its use, the affected workers and their representatives are informed.

*Only for public authorities and Union institutions; otherwise this paragraph is omitted.*

As a public authority or Union institution, **[YOUR_ORGANISATION_NAME]** verifies before use that the high-risk system is registered in the EU database and refrains from using it where it is not.

Where a high-risk system makes or supports decisions concerning natural persons, the persons concerned are informed that they are subject to its use. The channel and the location of that information are recorded per system in the AI register.

Where personal data are processed, the information provided by the provider is used to carry out or update the data protection impact assessment (Art. 26(9)). **[YOUR_ORGANISATION_NAME]** cooperates with the competent authorities in any action they take regarding a high-risk system.

Where a market surveillance authority requires a corrective action, a withdrawal, a recall or a restriction of making available for an AI system in use, the measure ordered is carried out within the period set by the authority; that period is in any event no later than fifteen working days, unless the Union harmonisation legislation provides otherwise. The authority, the reference and the deadline are recorded on the system concerned in the AI register, the implementation is run as a corrective action and its completion is evidenced there; until then the use of the system is halted.

## 11. Importer obligations for high-risk systems (Art. 23)

*Only where the organisation places high-risk AI systems from third countries on the market; otherwise this section is omitted.*

As an importer of high-risk AI systems from third countries, **[YOUR_ORGANISATION_NAME]** carries out the following obligations under Article 23 of the AI Regulation before such a system is placed on the market.

Before a system is placed on the market, it is verified that the provider has carried out the conformity assessment procedure and drawn up the technical documentation, that the system bears the CE marking and is accompanied by the EU declaration of conformity and the instructions for use, and that the provider has appointed an authorised representative.

Where there is reason to consider that a system is not in conformity, has been falsified or is accompanied by falsified documentation, it is not placed on the market until conformity is achieved. Where the system presents a risk to health, safety or fundamental rights, the provider and the market surveillance authority are informed.

The name, registered trade name or registered trade mark and contact address of **[YOUR_ORGANISATION_NAME]** are indicated on the system or, where that is not possible, on its packaging or accompanying documentation.

While a system is under the responsibility of **[YOUR_ORGANISATION_NAME]**, storage and transport conditions are chosen so that conformity is not jeopardised, where relevant.

A copy of the certificate issued by the notified body, of the instructions for use and of the EU declaration of conformity is kept at the disposal of the competent authorities for ten years after the system has been placed on the market. On reasoned request, all information and documentation needed to demonstrate conformity are provided to them, and **[YOUR_ORGANISATION_NAME]** cooperates in any action taken to address the risks posed by an imported system.

## 12. Distributor obligations for high-risk systems (Art. 24)

*Only where the organisation makes high-risk AI systems available on the market without being their provider or importer; otherwise this section is omitted.*

As a distributor of high-risk AI systems, **[YOUR_ORGANISATION_NAME]** carries out the following obligations under Article 24 of the AI Regulation before such a system is made available on the market.

Before a system is made available, it is verified that the system bears the CE marking, that it is accompanied by the EU declaration of conformity and the instructions for use, and that the provider and, where applicable, the importer have met their marking obligations.

Where there is reason to consider that a system is not in conformity, it is not made available until conformity is achieved. Where the system presents a risk, the provider or importer and the market surveillance authority are informed.

While a system is under the responsibility of **[YOUR_ORGANISATION_NAME]**, storage and transport conditions are chosen so that conformity is not jeopardised, where relevant.

Where it becomes apparent after a system has been made available that it is not in conformity, the corrective actions needed to bring it into conformity, withdraw it or recall it are taken, and the provider or importer and the competent authorities are informed. On reasoned request, all information and documentation needed to demonstrate conformity are provided to them, and **[YOUR_ORGANISATION_NAME]** cooperates in any action taken to address the risks.

## 13. Deemed provider (Art. 25)

*Only where one of the cases of Article 25 can occur; otherwise this section is omitted.*

Where **[YOUR_ORGANISATION_NAME]** places its own name or trademark on a third-party high-risk system, substantially modifies such a system, or changes its intended purpose so that it becomes high-risk, **[YOUR_ORGANISATION_NAME]** is considered its provider under Article 25(1) of the AI Regulation and carries out the provider obligations under Article 16 for that system. The original provider steps back for that system; the technical documentation and access needed are obtained from it (Article 25(2)).

Where **[YOUR_ORGANISATION_NAME]** places a high-risk AI system on the market together with its own product covered by the Union harmonisation legislation of Annex I, under its own name or trademark, **[YOUR_ORGANISATION_NAME]** is likewise considered the provider of that system under Article 25(3).

## 14. AI Literacy

Staff who use AI maintain a sufficient level of AI literacy for their role. AI literacy covers an understanding of the AI systems in use, their capabilities and limits, the risks they carry (such as incorrect output, bias and the disclosure of data) and the rules of this policy. This requirement rests on ISO/IEC 42001, Clause 7.2, under which the competence needed is determined, the required competence is ensured and evidence of it is retained. Art. 4 of the AI Regulation separately requires measures that support the development of AI literacy; the training and support provided under this policy are such measures.

The level of AI literacy is appropriate to the role and to the technical knowledge and experience of the person. Each member of staff completes the AI training assigned to their role, keeps it current and applies it in practice. The need for training is determined per role, and completion is recorded.

## 15. Consequences of Non-Compliance

A breach of this policy is treated in accordance with the disciplinary process of **[YOUR_ORGANISATION_NAME]**. The response is proportionate to the nature and gravity of the breach, to whether it was intentional or accidental, to whether the person had received adequate training, and to whether it is a first or a repeated occurrence. Depending on the severity, the response ranges from a documented discussion and refresher training, through a temporary restriction of access to AI services, to formal disciplinary measures where the breach is serious or wilful. Where a breach causes material damage or involves unlawful conduct, the organisation additionally pursues the remedies available to it under applicable law.

## 16. Roles and Responsibilities

- **All staff:** Use AI in accordance with this policy, use only approved AI services, keep confidential and personal data out of unapproved tools, verify AI output before using it, and report incidents and concerns.
- **Line managers and business units:** Report new or changed AI applications in their area, make sure their team is aware of and follows this policy, and support the oversight of the systems they use.
- **AI officer:** Maintains this policy and the AI inventory, assesses reported and shadow AI, and is the point of contact for the responsible use of AI. Handling incident reports rests with the person responsible for it under Section 7.
- **IT and approval of services:** Operates the approval process and the catalogue of approved IT and AI services, and provides the approved enterprise instances.
- **Data protection:** Advises on the handling of personal data in AI services and on the data protection requirements that a service meets before it is approved.

Where **[YOUR_ORGANISATION_NAME]** develops or operates its own AI systems, the staff involved additionally follow the provider duties set out in the AI Management System documents.

## 17. Reference Documents

- **AI Policy:** The overarching policy from which these rules derive.
- **Catalogue of approved IT and AI services:** The list of approved AI services and the conditions of their use, with the process to request a new service.
- **AI inventory:** The record of the AI systems in use and their classification, including the oversight arrangements for high-risk systems.
- **AI incident reporting:** The channel for reporting incidents and malfunctions of AI systems.
- **Information Security Policy and data protection documentation:** The rules that apply to the handling of information and personal data when using AI.

## 18. Review and Maintenance

This policy is reviewed:

- At least annually as part of the review cycle of the AI Management System.
- When new AI services are approved or new categories of AI use are introduced.
- After a significant AI incident or a pattern of incidents involving the use of AI.
- When changes to applicable law affect the rules of use, in particular new requirements of the AI Regulation or changes to its application stages.

Each review and each change is versioned and documented with a justification.
