Skip to content

Risk Operating Model ​

Risk in Modulos is a monetary value. The organization sets a shared taxonomy and a total risk appetite. Each AI System adds risks from the taxonomy and quantifies their threats. Quantify again when the AI System or its context changes.

For the expected-loss model and the reasons to use money, see Risk Quantification.

Where in Modulos ​

Modulos has risk work in two areas: the organization configuration and the Risks page of each AI System.

Organization level ​

Organization → Risk Management has seven tabs.

What you want to doWhere in ModulosOutput
See the total exposure, its trend, and the AI Systems and risks with the highest valuesOrganization → Risk Management → Risk OverviewOrganization totals
Find the groups, AI Systems, and risks that use the most of their limitsOrganization → Risk Management → Portfolio, Treemap viewA map of risk values against limits. See Treemap.
Compare the economic value of each AI System with its risk valueOrganization → Risk Management → Portfolio, Value and Risk viewAn assessment for each AI System. See Value and Risk.
Make the categories for totals and limitsOrganization → Risk Management → Category TaxonomyCategory structure
Make risks for all AI Systems, each with one category and its threat vectorsOrganization → Risk Management → Risk TaxonomyRisk library
Make threat vectors with one name and description for all AI SystemsOrganization → Risk Management → Threat Vector TaxonomyThreat vector library
Set the total risk appetite and the share of each categoryOrganization → Risk Management → Risk LimitsAppetite and category limits
Set how Modulos divides the appetite into AI System limitsOrganization → Risk Management → AI System Risk LimitsAI System risk limits

AI System level ​

What you want to doWhere in ModulosOutput
Add risks from the taxonomy and select their threat vectorsAI System → Risks → New RiskAI System risks and risk threats
Quantify one threatAI System → Risks → [Risk] → Threat Vectors → QuantifyA quantification run and a monetary value
Compare values with limitsAI System → Risks and Organization → Risk Management → Risk OverviewAI System and organization exposure

Who can do what ​

Organization roles

  • Risk Managers change the taxonomy and the risk limits.
  • Organization Admins and Members can open the Risk Management tabs. Without the Risk Manager role, they cannot change the taxonomy or the risk limits.
  • Organization Admins, Risk Managers, and Policy Managers have full access to the Risks page of each AI System.

AI System roles

  • Owners add, change, and delete risks, and quantify threats.
  • Editors add and change risks and their threat vectors. They cannot quantify threats or delete risks.
  • Viewers can see risks and quantification results.

Only users with a role in the AI System, or with full access to all AI Systems, see the Risks page of that AI System.

How risk becomes a portfolio signal ​

Quantification occurs at the level of the risk threat, which is one threat vector in one AI System risk:

  • An AI System risk gets one or more risk threats when you add it.
  • Each risk threat can have many quantification runs. The newest run sets the status and the value of the risk threat.
  • The value of a risk threat counts only while its newest run has the status Quantified. When the newest run is In Progress or Failed, the risk threat adds no value to the totals. This rule applies also when a previous run has the status Quantified.

Modulos adds the values from each risk threat up to the totals. A risk is the sum of its risk threats, and an AI System is the sum of its risks. The organization total is the sum of its AI Systems. A category total is the sum of the risks of that category in all AI Systems. See Platform objects and rollups.

  • A risk includes the value of each quantified risk threat, also when other risk threats of the risk are not quantified. The value of the risk and its limit use can increase when you quantify more of its risk threats.
  • The organization totals on the Risk Overview and Portfolio tabs include only AI Systems with the registration status Pending or Approved. They do not include the organization AI System.

Quantify again when the system, its data, its vendors, or its Controls change.

Risk limits ​

Risk limits divide the risk appetite of the organization into monetary limits for categories, AI Systems, and risks. A Risk Manager sets the appetite, the category limits, and the method for the AI Systems. Modulos calculates the limits of the AI Systems and of their risks.

Appetite and category limits ​

In Organization → Risk Management → Risk Limits:

  • Total Organization Risk Appetite is the maximum risk exposure of the organization across all AI Systems, as money.
  • Category Limit is the percentage of the appetite for one category. You can set the category limits only after the appetite has a value. Make the sum of the category limits 100%.
Risk Limits tab with the Total Organization Risk Appetite field, the Allocation used bar, and one Category Limit card for each risk category.

The Risk Limits tab sets the risk appetite and the share of each category.

  1. Total Organization Risk Appetite The maximum risk exposure for all AI Systems together, as money.
  2. Category Limit The share of the appetite for one category, as a percentage.
  3. Save Changes Keeps the appetite and the category limits.

AI System risk limits ​

In Organization → Risk Management → AI System Risk Limits, the AI System Risk Limit Methods card sets how Modulos divides the appetite into limits for the AI Systems:

  • Equal Distribution (default): each AI System gets the same limit.
  • Economic Value-Based Allocation: each AI System gets a limit in proportion to its annual economic value.

After you select a method, select Save Changes.

These rules apply to the two methods:

  • Only AI Systems with the registration status Pending or Approved get a limit. The organization AI System gets no limit.
  • A manual override gives one AI System a limit that does not change. To set it, select Edit in the AI Systems table, or use the Risk Management card in AI System → Settings → General Settings.
  • The AI Systems without an override divide the appetite that remains after the overrides.
  • Modulos calculates the limits again when an AI System goes into the status Pending or Approved, or out of it.
  • A Draft or Rejected AI System has no risk limit. A Rejected AI System and its risks show No risk limit. A Rejected AI System gets a limit again when it goes back to Pending.
AI System Risk Limits tab with the appetite card, the AI System Risk Limit Methods card, and the AI Systems table with Risk Limit and Economic Value columns.

The AI System Risk Limits tab sets how Modulos divides the appetite into limits for the AI Systems.

  1. Equal Distribution (default) Each AI System without an override gets the same share of the remaining appetite.
  2. Risk Limit The limit of the AI System. An icon shows a manual override.
  3. Edit Sets the economic value and a manual override for one AI System.

Risk limits in an AI System ​

For each category, the AI System limit multiplied by the category limit gives the category budget of the AI System. Modulos divides this budget equally between the risks of that category in the AI System. A manual override sets the limit of one risk. To set it, select Override with manual value in the Edit Risk dialog. The other risks of the category divide the remaining budget.

Warnings while you edit limits ​

Modulos shows an information banner in these conditions:

  • The sum of the category limits is not 100% (Risk Limits tab).
  • The sum of the AI System risk limits is not equal to the total appetite (AI System Risk Limits tab).
  • A manual override on a risk is not equal to the calculated limit (Edit Risk dialog).

Risk limits do not stop quantification. You can quantify a threat with all limit configurations. But the limit use and the limit lines in the charts are correct only when the limits agree.

A practical operating cadence ​

  • Set up one time: the taxonomy and the risk limits.
  • For each AI System: add the risks that apply, select their threat vectors, and quantify the most important threats first.
  • Continuously: decrease the threats with the highest expected loss through Controls, attach Evidence to the Controls, and quantify again after important changes.