Skip to content

AI System Settings ​

The settings of an AI System control its configuration after registration. They hold the registration details, the EU AI Act scoping, the user access, the frameworks and their versions, and the Sources.

What this is ​

In the AI System settings, you:

  • keep the scope statement and the registration details up to date
  • give AI System roles and keep separation of duties
  • control how frameworks apply and when they change to a new version
  • configure the Sources that tests read and that Scout uses as context

Where in Modulos ​

  • AI System → Settings

The settings have these tabs:

  • The General Settings tab: the cards Basic Information, AI System Details (not for the organization AI System), and, when risk management is on, Risk Management
  • The EU AI Act tab: the EU AI Act classification and scoping, when the AI System has the EU AI Act framework
  • The User Access tab: the AI System roles of the users
  • The Frameworks tab: Add Framework, the framework versions, updates, and the freeze of updates
  • The Sources tab: the Sources of the AI System, for tests and for Scout
  • The Danger Zone tab: the deletion of the AI System

The EU AI Act, Frameworks, and Sources tabs show when governance is on for your organization.

Who can do what ​

Permissions

  • The Settings item shows only for users who can change the AI System, give AI System roles, or delete the AI System. These users are the AI System Owners and the users with full access to all AI Systems: Organization Admins, Organization Risk Managers, and Organization Policy Managers.
  • Editors and Viewers do not see Settings. All members of the AI System can read the registration details and the EU AI Act classification on AI System → Overview.
General Settings tab of an AI System with the Basic Information, AI System Details, and Risk Management cards.

The General Settings tab holds the registration details and the risk limit of the AI System.

  1. General Settings Your rights and the features of your organization set which tabs show.
  2. Basic Information The name, the intended purpose, and the description, which is the scope statement of the AI System.
  3. AI System Details Current State, Target Go-Live Date, AI Lifecycle Stage, AI System Language, run scope, and Geographic Scope.
  4. Risk Management The economic value, the risk limit from the organization method, and the manual override.

How it works ​

General Settings ​

The General Settings tab has the registration details of the AI System. Click Save Changes to save them.

  • Basic Information: AI System Name, Intended Purpose (minimum 10 characters), and AI System Description. Write the description as the scope statement of the AI System.
  • AI System Details: Current State, Target Go-Live Date, AI Lifecycle Stage, AI System Language, Where will this AI System run?, and Geographic Scope. Current State and the run scope are required. The AI lifecycle stage helps you put the work and the reports into groups. It also sets which Controls the AI System dashboard counts: see AI lifecycle stage.

For the organization AI System, the tab has no Intended Purpose and no AI System Details card. Its AI System Language field is in Basic Information.

You can edit these details when the AI System is Pending, Approved, or Rejected. A Draft opens in the registration wizard, and you edit it there. See Register an AI System for the meaning of each field.

When risk management is on for your organization, the Risk Management card shows:

  • AI System Annual Economic Value: the value that the AI System gives in one year. Modulos compares it with the risk value in the Value and Risk view. With the Economic Value-Based Allocation method, it also sets the risk limit of the AI System
  • AI System Risk Limit: the limit that the organization method calculates from the risk limits of the organization (read-only)
  • Override with manual value: an Owner can set a Manual Override Limit when governance makes it necessary

A Rejected AI System has no risk limit, and the field shows No risk limit. See Registration approval.

EU AI Act ​

The EU AI Act tab shows for an AI System with the EU AI Act framework. On this tab, you record the classification that sets the scope of the AI System: which Requirements and Controls apply, and how auditors read them.

There are two methods to set the scope:

  • EU AI Act Questionnaire: a guided sequence of questions that finds the classification for you. Open it with the EU AI Act Questionnaire button on this tab. The Requirements page and the Controls page also have this button. Use it the first time that you scope an AI System. Also use it when your team must see the legal questions one by one.
  • Manual scoping: set each classification field on this tab. Use it when you know the classification, or when you change one field without the questionnaire.

The two methods write to the same configuration of the AI System and use the same scoping logic. You can change a scope that the questionnaire set by hand, and the opposite.

How the questionnaire runs ​

  1. Start: open the questionnaire. Modulos saves your answers while you work. You can close the dialog and continue at a different time.
  2. Questions: answer the questions about the system, your role, and its use.
  3. Result: a Summary shows the Requirements and Controls that your answers put out of scope.

The Confirm Scoping button applies the result. The listed Requirements and Controls go out of scope. Modulos attaches the Summary of Scoping as Evidence to the MCF-150 scoping Control. This makes the scoping decision auditable.

You can do the questionnaire again as many times as necessary. Re-take Questionnaire starts it again and deletes the answers that you gave. When you apply a new result, the scope changes.

Classification fields ​

The manual settings tab has eight classification fields in one grid. Their sequence is Product Category, Role, Use Case, Organization Establishment, Model Property, AI System Requirements, Prohibited AI Practices (Article 5), and Model Supply Terms. The two required fields are Product Category and Role. The table gives the fields in the sequence of the assessment. The EU AI Act Assessment card of the Overview uses the same sequence.

FieldWhat it records
Product Category (required)The type of AI in scope: AI System or General Purpose AI Model
Role (required)Your legal role in the value chain: Provider, Deployer, Importer, Distributor, or Authorised Representative. You can select more than one role.
Prohibited AI Practices (Article 5)The Article 5 result, with the one value Prohibited AI System. The assessment settles this field before the risk class.
Use CaseThe risk class that sets the Requirements in scope: High Risk, Annex III - Not High-Risk (an Annex III use case that is not high-risk because of Article 6(3)), Transparency, or Minimal Risk. Minimal Risk records that none of the other classes applies. It does not change the scope: the Requirements in scope are the same as with an empty Use Case.
Organization EstablishmentWhere the provider or deployer is established, for the territorial scope of Article 2. The values are Inside European Union and Outside of European Union.
Model PropertyA property of the model that changes the obligations: Systemic Risk
AI System RequirementsConditions on the AI System that turn more obligations on or off: Integration Support and Modification Support
Model Supply TermsThe supply terms of the model that change where the provider obligations are: Non-FOSS

Preview and apply ​

A manual change has two steps, a preview and an apply:

  1. Edit the classification fields and click Preview Changes.

  2. Examine the Preview Scope Changes dialog. It shows the number of items that change scope, for example "30 requirement(s) and 48 control(s) will change scope." It puts the items in four groups:

    • Requirements moving IN scope
    • Requirements moving OUT of scope
    • Controls moving IN scope
    • Controls moving OUT of scope

    Each item shows by its code. To see the name of a Requirement or Control, move the pointer on its code.

  3. Click Apply to make the change, or Cancel to keep the scope that you had.

The preview shows the effect of a classification change before you accept it. This is important while your team does the work, because Requirements and Controls that go out of scope can have Evidence attached.

When an applied change moves the scope, or records Minimal Risk, Modulos attaches a scoping summary as Evidence to MCF-150.

Record your reasoning

Use the AI System Classification Control (MCF-150) to record why you selected a classification. The tab shows this text with a link to the Control: "Explain and document your reasoning for your chosen EU AI Act Risk Classification in this control".

EU AI Act tab of an AI System with eight classification fields, the Preview Changes button, and the link to the MCF-150 Control.

The EU AI Act tab records the classification that sets the scope of the Requirements and Controls.

  1. EU AI Act Shows for an AI System with the EU AI Act framework.
  2. Product Category The eight classification fields. The fields Product Category and Role are required.
  3. Preview Changes Shows each Requirement and Control that goes in or out of scope before you apply the change.

User Access ​

On the User Access tab, AI System Owners give AI System roles. Add User to the AI System adds a user. The AI System roles are:

  • Owner: configures the AI System, gives and removes AI System roles, and manages all settings
  • Editor: does the work on Controls and attaches Evidence
  • Viewer: reads the AI System and adds comments for assurance, but cannot change items

A user can have more than one role on the same AI System.

Multiple Owners ​

An AI System can have more than one Owner at the same time, and it must have minimum one Owner. Modulos does not let you remove the last Owner. The message is "You cannot remove the only owner of the AI System." Owners, Organization Admins, and Super Admins can add or remove the Owner role. The log of the AI System records each Owner change.

Use more than one Owner when an AI System must have continuous coverage, for example during vacations or role changes. Also use more than one Owner when the accountability is in more than one team, for example Product and Compliance. With two Owners, you can also keep the Owner who does the work apart from the Owner who examines the Requirements.

The AI Registry shows all Owners of each AI System. When the organization deactivates a user, that user stays in the Owners of the AI System, with a sign for deactivated users. This keeps the audit trail complete. That role assignment gives no more access. Add a new Owner before you deactivate a user. You can add only activated users as members or Owners.

User Access tab of an AI System with two Owners, two Editors, one Viewer, and the Add User to the AI System button.

The User Access tab shows the AI System roles of each user. An AI System can have more than one Owner.

  1. Owner An AI System can have more than one Owner and must have minimum one.
  2. Edit Changes the AI System roles of the user.
  3. Add User to the AI System Adds an activated user of the organization with one or more AI System roles.

Frameworks and versioning ​

The frameworks set the compliance scope of an AI System, and each framework has versions. The Frameworks tab shows the frameworks of the AI System with their Current Version and Latest Version. Click a framework to open the change log of its version. Check all the change logs in that dialog opens the framework in the Frameworks library.

On this tab:

  • Add Framework adds a framework. The button is not available while a framework is not at its latest version: update the frameworks that are not current first. When updates are frozen, the button does not show.
  • Update Frameworks changes the outdated frameworks to their latest versions.
  • Freeze Updates shows only to Organization Admins, Organization Risk Managers, and Organization Policy Managers, not to AI System Owners.

See Frameworks in Modulos for what an update changes and what a freeze does.

Frameworks tab of the AI System settings with the frameworks table, Current Version and Latest Version columns, the Freeze Updates toggle, and the Update Frameworks and Add Framework buttons.

On the Frameworks tab, you add frameworks and control when the AI System accepts framework updates.

  1. Frameworks The frameworks of the AI System, with their versions.
  2. Freeze Updates Stops all updates and additions of frameworks. Only Organization Admins, Risk Managers, and Policy Managers see it.
  3. Update Frameworks Changes the frameworks to their latest versions.
  4. Current Version Compare it with the Latest Version column.
  5. Add Framework Adds a framework. Not available while a framework is not at its latest version.

Sources vs Connectors ​

The Sources tab shows the Sources of the AI System. A Source is a service account that you attach to an AI System. Tests read it, and Scout uses it as context about the AI System.

A Connector is not on this tab: it belongs to one user, who adds it in User Settings → Connectors. See Sources and Connectors.

Delete an AI System ​

The Danger Zone tab shows to Owners and to the users with full access to all AI Systems. On this tab, Delete AI System deletes a Pending, Approved, or Rejected AI System after a confirmation. A Draft has no Settings. To delete a Draft, see Delete a Draft.

How to use it ​

  • Keep the scope current: keep a clear description that agrees with what you govern.
  • Give roles carefully: use Owners, Editors, and Viewers to keep separation of duties.
  • Control framework changes: update frameworks when changes are possible, and have an Organization Admin freeze updates before the audit.
  • Configure Sources: connect the AI System to the systems that its tests and Scout read.
  • Use the EU AI Act settings when they apply: record the classification that sets the Requirements and Controls in scope.

Important considerations ​

  • Sources hold credentials with high trust. Use service accounts with the minimum privileges, and change the credentials when the ownership changes.
  • If you do not see a tab on this page, the cause can be your rights or the features of your organization.
  • The deletion of an AI System is permanent. Export it first if you must keep an audit trail.