Zum Hauptinhalt springen
AI Compliance Starter Kit · Template

AI Communication Plan

Updated on 2 min Reviewed by: Cenedril Editorial
ISO/IEC 42001 7.4EU AI Act Art. 26, 50, 73

Who is told, and when, that a chatbot answers in customer service? Who reports a serious incident to the market surveillance authority, and within which deadline? ISO/IEC 42001 requires in clause 7.4 that the need for communication about the AI management system is determined; the AI Act makes part of it mandatory. The AI communication plan records in one table what goes to whom, when and through which channel, and who is responsible for it.

Who is the plan for?

For any organisation with an AI management system under ISO/IEC 42001. Deployers delete the rows that concern providers only; providers of high-risk AI systems also cover the communication under Art. 17(1)(j) with the plan.

What goes into the plan?

  • Purpose: need for internal and external communication, statutory communication duties, fixed contact channel, accessibility
  • Communication events: table with subject, frequency, channel, audience, documentation needed and the responsible person
  • Content per event: key messages for each of the 16 rows
  • Review and basis: triggers for review, standards and legal basis

Template: AI Communication Plan

Full policy text

AI Communication Plan

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 plan sets out, under ISO/IEC 42001 clause 7.4, what is communicated about the AI management system, when, to whom, through which channel and by whom. The table contains the statutory communication duties of Regulation (EU) 2024/1689 and the internal communication as default rows. Rows that concern providers only (authority requests, supply chain, customer information, decommissioning, reporting channel for deployers, fundamental rights authorities) are omitted for deployers alone. The column "Documentation needed" is a proposal under ISO/IEC 42001 A.6.2.7, which the organisation determines per audience.

1. Purpose

ISO/IEC 42001:2023 Cl. 7.4 requires [YOUR_ORGANISATION_NAME] to determine the internal and external communications relevant to the AI management system. This plan sets out what is communicated, when, to whom, through which channel and who is responsible.

With AI, part of the communication is required by law. Regulation (EU) 2024/1689 (AI Act) requires transparency information for people who interact with an AI system or are affected by its output (Art. 50), information to workers and their representatives before a high-risk AI system is put into use at the workplace (Art. 26(7)), and reporting of serious incidents to the market surveillance authority (Art. 73). These duties sit in this plan alongside voluntary internal communication.

Where a system's conformity is assessed by a notified body under Annex VII, changes to the system that are relevant to its conformity are reported to the issuing notified body for assessment before they take effect (Annex VII point 4.7); this reporting is part of this plan.

Affected persons, deployers and other interested parties can reach [YOUR_ORGANISATION_NAME] about its AI systems at any time through the AI reporting channel ([AI_REPORTING_CHANNEL, e.g. reporting form or mailbox]). Complaints, indications of risk and requests for explanation arrive there and are handled as an AI observation or a request for explanation.

Communication to external parties and to staff is accessible: content is understandable without colour alone, reachable with assistive technology and written in plain language; EN 301 549 applies to digital channels. Form and language follow the age and needs of the audience, for example children, older people or persons with disabilities, and are adapted on request.

The communication plan is maintained as a living document and reviewed at least annually or whenever the AI management system changes materially.

2. Communication events

# Subject Frequency Channel Audience Documentation needed Responsible
1 AI policy: annual communication to all employees Annually Email All employees AI policy [RESPONSIBLE]
2 AI usage rules: announced on introduction and on every change On occurrence Intranet All employees AI usage policy [RESPONSIBLE]
3 Transparency information for people interacting with the AI system (Art. 50) On occurrence AI system or product interface Affected persons, users of the system – [RESPONSIBLE]
4 Information to workers before putting a high-risk AI system into use (Art. 26(7)) On occurrence Meeting or presentation Affected workers and their representatives – [RESPONSIBLE]
5 Reporting serious incidents to the market surveillance authority or the AI Office and to the provider (Art. 73, Art. 75(1a)) On occurrence Notification to authority Market surveillance authority or, where the AI Office is exclusively competent under Art. 75(1), the AI Office; provider of the system Incident report [RESPONSIBLE]
6 Notifying users and affected people after an AI incident On occurrence AI system or product interface Affected people, customers, partners – [RESPONSIBLE]
7 Informing users about changes to an AI system On occurrence AI system or product interface Users of the changed system, person holding oversight Instructions for use (Art. 13) [RESPONSIBLE]
8 AI literacy: annual training for people operating or using AI systems (ISO/IEC 42001 Cl. 7.2, AI Act Art. 4) Annually Training or workshop Everyone operating or using AI systems Training material [RESPONSIBLE]
9 Status of AI objectives and risk situation: report to top management Annually Meeting or presentation Top management – [RESPONSIBLE]
10 AI system withheld or suspended: informing the parties concerned (Art. 23(2), Art. 26(5), Art. 24(2)) On occurrence Email Provider or importer, affected workers and customers, and the market surveillance authority where there is a risk – [RESPONSIBLE]
11 Information and documents on a reasoned request from an authority (Art. 21) On occurrence Notification to authority Competent national authority, notified body Technical documentation (Annex IV), EU declaration of conformity [RESPONSIBLE]
12 Informing distributors, deployers and authorities on non-conformity (Art. 20) On occurrence Email Distributors, deployers, importers, authorised representatives, competent authorities – [RESPONSIBLE]
13 Capabilities, limits of the validated domain and division of responsibility On occurrence Product documentation or terms of use Customers and downstream providers Instructions for use (Art. 13) [RESPONSIBLE]
14 Notice of the decommissioning of a system On occurrence Email Deployers, distributors and importers – [RESPONSIBLE]
15 Reporting of serious incidents and risks by deployers On occurrence AI system or product interface Deployers of the provided systems Instructions for use (Art. 13) [RESPONSIBLE]
16 Access to documentation on request of fundamental rights authorities and bodies providing access to data (Art. 77) On occurrence Notification to authority Authorities protecting fundamental rights Technical documentation (Annex IV) [RESPONSIBLE]

3. Content of the individual communication events

The sections below record the key messages for each event.

AI policy: annual communication to all employees

  • Purpose of the AI management system and top management's commitment
  • Principles for handling AI in the organisation
  • Scope: which areas and systems are covered
  • Where the AI policy can be viewed

AI usage rules: announced on introduction and on every change

  • Which AI tools are approved and which are not
  • Handling confidential and personal data in prompts
  • Duty to review AI output before using it
  • Reporting route for unusual or faulty output

Transparency information for people interacting with the AI system (Art. 50)

  • Notice that this is an AI system
  • Marking of AI-generated content where required
  • Purpose of the system and type of input processed
  • Contact point for questions and complaints
  • Timing: the notice is given before or at the first interaction

Information to workers before putting a high-risk AI system into use (Art. 26(7))

  • Which system is used and for what purpose
  • Which decisions it prepares or supports
  • Who exercises human oversight
  • Timing: workers are informed before the system is put into use

Reporting serious incidents to the market surveillance authority or the AI Office and to the provider (Art. 73, Art. 75(1a))

  • Description of the incident and the persons affected
  • AI system concerned, the organisation's role and the system identifier
  • Immediate measures taken
  • Deadlines under Art. 73: immediately, at the latest 15 days after becoming aware; 2 days for widespread infringements or disruption of critical infrastructure; 10 days in the event of a death

Notifying users and affected people after an AI incident

  • Type of incident and the AI system concerned
  • Possible severity for those being notified
  • Immediate measures taken and what of them is noticeable
  • Contact point for questions and complaints
  • Time of the notification and whether another one follows

Informing users about changes to an AI system

  • Which AI system was changed and what changes in its behaviour
  • What the change means for day-to-day use
  • Whether outputs need to be checked differently than before
  • From when the change takes effect
  • Where to raise questions

AI literacy: annual training for people operating or using AI systems (ISO/IEC 42001 Cl. 7.2, AI Act Art. 4)

  • What AI systems can do and where their limits are
  • Recognising faulty or biased output
  • Legal frame: prohibitions, risk classes and the duties of the organisation's own role
  • How to act on anomalies and incidents

Status of AI objectives and risk situation: report to top management

  • Status of the AI objectives and reported results
  • Changes in the risk situation since the last assessment
  • New or retired AI systems and their classification
  • Open obligations and how far they have been implemented

AI system withheld or suspended: informing the parties concerned (Art. 23(2), Art. 26(5), Art. 24(2))

  • Which status was set: withheld before making the system available, or its use suspended, with the reason
  • The AI system concerned, with its identifier, and the time of the measure
  • Who is informed: the provider or importer, internally the affected workers, and, where the system is used towards customers, the customers concerned
  • Where there is a risk to health, safety or fundamental rights, also the competent market surveillance authority, without delay

Information and documents on a reasoned request from an authority (Art. 21)

  • Package to hand over: technical documentation, EU declaration of conformity, evidence of the conformity assessment, monitoring plan
  • Access to the automatically generated logs in so far as they are under the provider’s control
  • A language version the requesting authority readily understands
  • Point of entry and the role responsible for answering

Informing distributors, deployers and authorities on non-conformity (Art. 20)

  • Which step was taken: brought into conformity, withdrawn, disabled or recalled
  • Systems and batches concerned, with their identifiers
  • Where there is a risk to health, safety or fundamental rights, additionally the market surveillance authorities of the countries where the system was made available

Capabilities, limits of the validated domain and division of responsibility

  • Capabilities and intended purpose of the system as delivered
  • Limits of the validated domain: what the system is tested for and what it is not
  • Uses outside those limits and what applies then
  • Measured accuracy per domain from the technical documentation
  • Division of responsibility between provider and customer, including the duties that stay with the customer
  • Feedback and contact route for anomalies, questions and complaints

Notice of the decommissioning of a system

  • From when the system is no longer provided or supported
  • What the deployer does: end use, secure its own data, inform its own users
  • How data, logs and models are handled
  • Who can be contacted with questions

Reporting of serious incidents and risks by deployers

  • Reporting route to the provider as named in the instructions for use
  • What a report contains: system, version, time, course of events, affected persons
  • When the deployer suspends use (Art. 26(5))
  • Acknowledgement of receipt and next steps

Access to documentation on request of fundamental rights authorities and bodies providing access to data (Art. 77)

  • Which authority requests and on which basis (Art. 77)
  • Which documents are provided, in a language the authority readily understands
  • Deadline of the authority or the deadline set on receipt
  • Confidentiality under Art. 78

4. Review and maintenance

The communication plan is reviewed at least annually as part of the management review. Additional reviews are triggered by:

  • A new AI system entering the register, or a change to its risk class
  • A change of role towards an AI system, for example from deployer to provider
  • Changes to the scope of the AI management system or to the organisational structure
  • Serious incidents and complaints concerning AI systems
  • Results of internal or external audits

5. Standards and legal basis

  • ISO/IEC 42001:2023, Cl. 7.4 (Communication)
  • ISO/IEC 42001:2023, Cl. 5.2 (AI policy, communicated within the organisation)
  • Regulation (EU) 2024/1689, Art. 50 (Transparency obligations)
  • Regulation (EU) 2024/1689, Art. 26(7) (Information to workers)
  • Regulation (EU) 2024/1689, Art. 73 (Reporting of serious incidents)
  • Regulation (EU) 2024/1689, Annex VII point 4.7 (report of conformity-relevant changes to the notified body)

Sources

Frequently asked questions

Does a deployer have to report serious incidents itself?

A deployer that identifies a serious incident immediately informs first the provider, and then the importer or distributor and the market surveillance authority (Art. 26(5)). The row on reporting serious incidents therefore names the authority and the provider as its audience.

What goes into the column "Documentation needed"?

ISO/IEC 42001 requires in A.6.2.7 that the system documentation needed by each relevant category of interested parties is determined. The template proposes the obvious documents, such as the instructions for use for customers or the technical documentation for authorities; the organisation checks and completes the column.

How often is the plan reviewed?

At least annually in the management review, and additionally for a new AI system, a change of role, a changed scope, after serious incidents and after audits.

No Cookies!

This wiki collects nothing, bakes nothing and leaves nothing behind. There's nothing to consent to. Privacy doesn't get better than this.