diff --git a/.gitignore b/.gitignore index 7bd12fef76..de8667d222 100644 --- a/.gitignore +++ b/.gitignore @@ -35,3 +35,7 @@ site/validmind-docs.yaml # Python API docs are now generated on the fly site/validmind + +# Cursor rules +.cursor/rules/ +.cursor/skills/ \ No newline at end of file diff --git a/site/training/ai-governance/ai-governance-lifecycle.svg b/site/training/ai-governance/ai-governance-lifecycle.svg new file mode 100644 index 0000000000..2a15b5e625 --- /dev/null +++ b/site/training/ai-governance/ai-governance-lifecycle.svg @@ -0,0 +1,89 @@ + + + + + + + + + + + AI governance lifecycle: Intake, Assessment, Documentation, Validation, Approval, Deployment, Monitoring, Review + + + + + + 1. Intake + + Use case owner + Registers AI system + + + + + + 2. Assessment + + Governance / Risk + Classify & assess + + + + + + 3. Documentation + + Owner / Risk + Document model & use case + + + + + + 4. Validation + + Validator + Test & validate model + + + + + + 5. Approval + + Committee / Compliance + Sign-off + + + + + + 6. Deployment + + Owner / IT + Deploy to production + + + + + + 7. Monitoring + + Risk / Ops + Ongoing oversight + + + + + + 8. Review + + Governance / Audit + Periodic review + + + + + Re-assess / re-approve + diff --git a/site/training/ai-governance/ai-governance-mrm-overlap.svg b/site/training/ai-governance/ai-governance-mrm-overlap.svg new file mode 100644 index 0000000000..70ee4f5c04 --- /dev/null +++ b/site/training/ai-governance/ai-governance-mrm-overlap.svg @@ -0,0 +1,11 @@ + + AI governance and MRM as parallel use cases with overlap + + + + + + + AI Governance + MRM + diff --git a/site/training/ai-governance/ai-governance-register.qmd b/site/training/ai-governance/ai-governance-register.qmd index d910c2b28e..00ec6934df 100644 --- a/site/training/ai-governance/ai-governance-register.qmd +++ b/site/training/ai-governance/ai-governance-register.qmd @@ -52,7 +52,7 @@ Learn how to use {{< var vm.product >}} for **AI governance** to establish organ ::: {.attn} ## In this course -:::: {.flex .flex-wrap .justify-around} +:::: {.flex .flex-wrap .justify-around .mt4} ::: {.w-50-ns .pr3} #### Total duration diff --git a/site/training/ai-governance/analytics-dashboard.png b/site/training/ai-governance/analytics-dashboard.png new file mode 100644 index 0000000000..46a0a1b80a Binary files /dev/null and b/site/training/ai-governance/analytics-dashboard.png differ diff --git a/site/training/ai-governance/analytics-risk-overview.png b/site/training/ai-governance/analytics-risk-overview.png new file mode 100644 index 0000000000..ef1b75d87c Binary files /dev/null and b/site/training/ai-governance/analytics-risk-overview.png differ diff --git a/site/training/ai-governance/compliance-and-oversight.qmd b/site/training/ai-governance/compliance-and-oversight.qmd index dd6407eaac..6b58efc5c9 100644 --- a/site/training/ai-governance/compliance-and-oversight.qmd +++ b/site/training/ai-governance/compliance-and-oversight.qmd @@ -15,7 +15,7 @@ format: help: true controls-back-arrows: visible transition: slide - theme: [default, ../assets/slides.scss, ./slides.scss] + theme: [../assets/slides.scss, ../assets/use-cases.scss] slide-number: true chalkboard: false preview-links: auto @@ -50,24 +50,26 @@ This fourth module is part of a four-part series: 1. [Regulatory mapping](#regulatory-mapping) 2. [Dashboards and reporting](#dashboards-and-reporting) 3. [Audit evidence](#audit-evidence) -4. [Ongoing compliance](#ongoing-compliance) +4. [Continuous compliance](#continuous-compliance) ::: {{< include /training/assets/_revealjs-navigation.qmd >}} -# Regulatory mapping {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +# Regulatory mapping {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Understanding requirements {.center} -Key AI governance regulations include: +Key governance regulations include: +::: {.f2 .mt4} | Regulation | Jurisdiction | Focus | |------------|--------------|-------| -| EU AI Act | European Union | Risk-based AI regulation | -| SR 11-7 | United States | Model risk management | -| SS1/23 | United Kingdom | MRM principles | -| E-23 | Canada | Enterprise-wide MRM | +| [EU AI Act](https://eur-lex.europa.eu/eli/reg/2024/1689){target="_blank"} | European Union | Risk-based AI regulation | +| [SR 11-7](https://www.federalreserve.gov/supervisionreg/srletters/sr1107.htm){target="_blank"} | United States | Model risk management | +| [SS1/23](https://www.bankofengland.co.uk/prudential-regulation/publication/2023/may/model-risk-management-principles-for-banks-ss){target="_blank"} | United Kingdom | MRM principles | +| [E-23](https://www.osfi-bsif.gc.ca/en/guidance/guidance-library/guideline-e-23-model-risk-management-2027){target="_blank"} | Canada | Enterprise-wide MRM & AI/ML risk | +::: ## Mapping to {{< var vm.product >}} {.center} @@ -76,116 +78,227 @@ Each regulation maps to platform capabilities: - **Inventory** — Registration and classification requirements - **Documentation** — Technical documentation articles - **Workflows** — Approval and oversight processes -- **Monitoring** — Ongoing compliance requirements +- **Monitoring** — Continuous compliance requirements + +## [Implementation guides]{.hidden} {.center} + +:::: {.columns} + +::: {.column width="50%" .pl3 .pr3} + +::: {.pl4 .mb4} +[Put Governance into Practice]{.f1 .b style="color: #DE257E;"} + +::: {.f3} +Use our implementation guides for: +::: + +::: {.f4 .pl3 .pt2} +[[AI Governance
[[[EU AI ACT]{.green}]{.bubble}]{.f5}](/about/use-cases/ai-governance.qmd){target="_blank"}]{.button .shadow-5-ns .f4 .content-tile style="width: 100%;"} + +[[Model Risk Management
[[[SR 11-7]{.green}]{.bubble} [[SS1/23]{.green}]{.bubble} [[E-23]{.green}]{.bubble}]{.f5}](/about/use-cases/model-risk-management.qmd){target="_blank"}]{.button .shadow-5-ns .f4 .content-tile style="width: 100%;"} +::: +::: + +::: + +::: {.column width="50%" .pl5 .pr2 .tilt-left} + +::: {.pl4 .mb4 .bg-dark-green .white .ba .pa3 .br3 .shadow-5-ns .o-90} +[{{< fa lightbulb >}} What these guides cover]{.f3 .b} + +::: {.f4} +- Inventory setup, custom fields, risk classification +- Documentation templates aligned to frameworks +- Automated testing for data quality and accuracy +- Independent validation, effective challenge, remediation +- Ongoing monitoring and drift detection +- Governance roles and approval workflows +- Audit trails for compliance evidence +::: +::: -## Implementation guides {.center} +::: +:::: + -# Dashboards and reporting {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +# Dashboards and reporting {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Governance dashboards {.center} +:::: {.columns} + +::: {.column width="50%" .pr3} Dashboards provide visibility into: +::: {.f2} - AI system portfolio overview - Risk distribution across tiers - Workflow status and bottlenecks - Compliance metrics +::: +::: + +::: {.column width="50%" .pl3 .mt5 .f3} +![Analytics dashboard](analytics-dashboard.png){fig-alt="A screenshot showing the ValidMind Analytics view with portfolio overview and compliance metrics" .screenshot .tilt-left} +::: + +:::: ## Configuring dashboards {.center} -In {{< var vm.product >}}: +The **Analytics** section provides: + +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Pre-built dashboards for governance oversight + +[{{< fa check >}}]{.dark-pink}   Customizable widgets for key metrics -1. Navigate to **Analytics** -2. Select or create dashboard views -3. Configure widgets for key metrics -4. Set up automatic refresh and sharing +[{{< fa check >}}]{.dark-pink}   Filters by risk tier, status, and ownership + +[{{< fa check >}}]{.dark-pink}   Export options for board reporting +::: ## Reporting {.center} Generate reports for: -- Board and executive oversight -- Regulatory submissions -- Internal audit -- Compliance monitoring +:::: {.flex .flex-wrap .justify-around .mt4} -# Audit evidence {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +::: {.w-40-ns .content-block-teal .mb3} +**Executive oversight** +::: + +::: {.w-40-ns .content-block-teal .mb3} +**Regulatory submissions** +::: + +::: {.w-40-ns .content-block-teal .mt3} +**Internal & external audits** +::: + +::: {.w-40-ns .content-block-teal .mt3} +**Compliance monitoring** +::: + +:::: + +# Audit evidence {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Types of audit evidence {.center} {{< var vm.product >}} automatically captures: +::: {.f2 .mt4} | Evidence type | Description | |--------------|-------------| | Registration records | AI system registration and updates | | Approval history | Who approved what, when | | Documentation | Technical and governance documentation | | Change history | All modifications with timestamps | +::: ## Generating evidence {.center} To prepare for audits: -1. Export documentation and reports -2. Generate activity logs -3. Produce compliance summaries -4. Package evidence for auditors +1. **Export** documentation and reports +2. **Generate** activity logs +3. **Produce** compliance summaries +4. **Package** evidence for auditors ## Maintaining audit readiness {.center} Best practices: -- Keep documentation current -- Complete workflows promptly -- Document decisions and rationale -- Perform periodic self-assessments +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Keep documentation current -# Ongoing compliance {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +[{{< fa check >}}]{.dark-pink}   Complete workflows promptly -## Continuous monitoring {.center} +[{{< fa check >}}]{.dark-pink}   Document decisions and rationale -Ongoing compliance requires: +[{{< fa check >}}]{.dark-pink}   Perform periodic self-assessments +::: + +# Continuous compliance {background-image="../assets/digitalspace-colours.svg" .title-accent} + +## Ongoing monitoring {.center} + +:::: {.columns} + +::: {.column width="50%" .f2 .pr3} +Continuous compliance requires: - Regular review of AI systems - Performance monitoring - Change tracking - Issue management +::: + +::: {.column width="50%" .pl3 .tr .f3 .nt3 .tilt-left} +![Ongoing monitoring documentation](../assets/vm-ongoing-monitoring.gif){width="80%" fig-alt="An animation of ongoing model monitoring being scrolled through on the ValidMind Platform showing performance metrics and potential issues" .screenshot} + +![Identify potential issues early](../assets/target-prediction-distribution-plot-zoomed.png){width="80%" fig-alt="A screenshot of a graph that shows potential data drift over time" .screenshot} +::: + +:::: ## Periodic reviews {.center} +:::: {.columns} + +::: {.column width="55%" .pr3} Configure periodic reviews: -1. Set review frequencies by risk tier -2. Assign reviewers -3. Track completion -4. Document review outcomes +- Set review frequencies by risk tier +- Assign reviewers +- Track completion +- Document review outcomes + +::: + +::: {.column width="45%" .pl3 .tr .nt3 .tilt-left .f3} +![Periodic review workflow](workflows-periodic-review.png){fig-alt="A screenshot showing the ValidMind Periodic Review workflow with frequency, reviewer assignments, completion tracking, and document outcomes" .screenshot} +::: + +:::: ## Staying current {.center} -AI regulations evolve. Stay current by: +AI regulations **evolve**. Stay current by: -- Monitoring regulatory updates -- Updating classification criteria -- Revising documentation templates -- Training staff on changes +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Monitoring regulatory updates + +[{{< fa check >}}]{.dark-pink}   Updating classification criteria + +[{{< fa check >}}]{.dark-pink}   Revising documentation templates + +[{{< fa check >}}]{.dark-pink}   Training staff on changes +::: ## Course completion {.center} -Congratulations on completing **AI Governance**! +Congratulations on completing **AI Governance**! You have learned: -You have learned: +::: {.mt4 .f2 .mb5} +[{{< fa check >}}]{.dark-pink}   AI governance concepts and terminology -- AI governance concepts and terminology -- How to manage AI use cases -- Workflow configuration for governance -- Compliance and oversight practices +[{{< fa check >}}]{.dark-pink}   How to manage AI use cases + +[{{< fa check >}}]{.dark-pink}   Workflow configuration for governance + +[{{< fa check >}}]{.dark-pink}   Compliance and oversight practices +::: ::: {.tc} [Return to AI Governance](ai-governance-register.qmd){.button target="_blank"} diff --git a/site/training/ai-governance/configuring-ai-workflows.qmd b/site/training/ai-governance/configuring-ai-workflows.qmd index 8e337f1d56..2c126076cf 100644 --- a/site/training/ai-governance/configuring-ai-workflows.qmd +++ b/site/training/ai-governance/configuring-ai-workflows.qmd @@ -15,7 +15,7 @@ format: help: true controls-back-arrows: visible transition: slide - theme: [default, ../assets/slides.scss, ./slides.scss] + theme: [../assets/slides.scss, ../assets/use-cases.scss] slide-number: true chalkboard: false preview-links: auto @@ -56,107 +56,147 @@ This third module is part of a four-part series: {{< include /training/assets/_revealjs-navigation.qmd >}} -# Intake workflows {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +# Intake workflows {background-image="../assets/digitalspace-colours.svg" .title-accent} ## What is an intake workflow? {.center} -Intake workflows control how new AI systems enter your governance program. +Intake workflows **control** how new AI systems enter your governance program. They ensure: -They ensure: +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   All AI systems are registered -- All AI systems are registered -- Required information is captured -- Initial classification is performed -- Ownership is assigned +[{{< fa check >}}]{.dark-pink}   Required information is captured + +[{{< fa check >}}]{.dark-pink}   Initial classification is performed + +[{{< fa check >}}]{.dark-pink}   Ownership is assigned +::: ## Configuring intake {.center} In {{< var vm.product >}}: -1. Define required fields for registration -2. Set up validation rules -3. Configure automatic notifications -4. Route new registrations for review +1. **Define** required fields for registration +2. **Set up** validation rules +3. **Configure** automatic notifications +4. **Route** new registrations for review + +## [Example: Risk classification workflow]{.f2} {.center .tc} + +::: {.tc} +![{{< fa magnifying-glass >}} [Click to enlarge]{.f3}](risk-classification-workflow.png){width="35%" fig-alt="A screenshot showing a Risk Classification Workflow with model stage changes, user actions, approvals, and condition branches for different risk levels" .screenshot} +::: + +## Risk classification workflow steps {.center} -## Hands-on: Review workflow configuration {.center} +::: {.f2} +This example intake workflow demonstrates key AI governance patterns: +::: -1. Navigate to **Settings > Workflows** -2. Explore the available workflow templates -3. Identify which workflows would support intake for AI systems +::: {.f3} +- **Model Stage Change** — Sets model stage to [In Risk Classification]{.bubble} +- **User Action v2** — Collects risk classification materials with deadline +- **Broadcast** — Notifies stakeholders of pending risk rating +- **Approval** — Routes to Model Validation Governance for review +- **Condition Branch** — Different paths for High Risk, Medium Risk, or Default +::: -# Approval workflows {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +# Approval workflows {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Purpose of approval workflows {.center} -Approval workflows ensure AI systems receive appropriate review before deployment. +Approval workflows **ensure** AI systems receive appropriate review before deployment. They implement: -They implement: +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Risk-based approval gates -- Risk-based approval gates -- Multiple approver requirements -- Segregation of duties -- Documented decision-making +[{{< fa check >}}]{.dark-pink}   Multiple approver requirements + +[{{< fa check >}}]{.dark-pink}   Segregation of duties + +[{{< fa check >}}]{.dark-pink}   Documented decision-making +::: ## Designing approval flows {.center} Consider these design principles: +::: {.f2 .mt4} | Risk level | Approval requirements | |------------|----------------------| | High-risk | Multiple approvers, governance committee | | Medium-risk | Manager + compliance review | | Low-risk | Single approver | +::: ## Configuring approvals {.center} +:::: {.columns} + +::: {.column width="50%" .pr3} {{< var vm.product >}} supports: - Workflow states and transitions - Role-based approval authorities - Sequential and parallel approvals - Approval deadline tracking +::: -# Human oversight controls {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +::: {.column width="50%" .pl3 .f3 .mt3} +![Approval step configuration](workflow-approval-step-configuration.png){fig-alt="A screenshot showing approval step configuration with role selection in the workflow builder" .screenshot .tilt-left} +::: + +:::: + +# Human oversight controls {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Regulatory context {.center} -The EU AI Act (Article 14) requires human oversight for high-risk AI systems. +The EU AI Act (Article 14) **requires** human oversight** for high-risk AI systems. Human oversight includes: + +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Ability to understand AI outputs -Human oversight includes: +[{{< fa check >}}]{.dark-pink}   Capability to intervene -- Ability to understand AI outputs -- Capability to intervene -- Authority to override decisions -- Mechanisms for feedback +[{{< fa check >}}]{.dark-pink}   Authority to override decisions + +[{{< fa check >}}]{.dark-pink}   Mechanisms for feedback +::: ## Implementing oversight {.center} Use {{< var vm.product >}} to document and enforce oversight: -- Record oversight requirements in documentation -- Configure review workflows for AI outputs -- Track intervention decisions -- Maintain evidence of human involvement +- **Record** oversight requirements in documentation +- **Configure** review workflows for AI outputs +- **Track** intervention decisions +- **Maintain** evidence of human involvement ## Review workflows {.center} Set up periodic review workflows: -1. Define review frequency by risk tier -2. Configure reviewer assignments -3. Track review completion -4. Escalate overdue reviews +1. **Define** review frequency by risk tier +2. **Configure** reviewer assignments +3. **Track** review completion +4. **Escalate** overdue reviews -# Escalations {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +# Escalations {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Escalation paths {.center} Configure escalations for: -- Overdue approvals -- High-risk findings -- Policy violations -- Compliance incidents +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Overdue approvals + +[{{< fa check >}}]{.dark-pink}   High-risk validation issues + +[{{< fa check >}}]{.dark-pink}   Policy violations + +[{{< fa check >}}]{.dark-pink}   Compliance incidents +::: Escalation ensures issues reach appropriate decision-makers. diff --git a/site/training/ai-governance/custom-fields.gif b/site/training/ai-governance/custom-fields.gif new file mode 100644 index 0000000000..489cc272f5 Binary files /dev/null and b/site/training/ai-governance/custom-fields.gif differ diff --git a/site/training/ai-governance/governance-pillars.svg b/site/training/ai-governance/governance-pillars.svg new file mode 100644 index 0000000000..a654b3a974 --- /dev/null +++ b/site/training/ai-governance/governance-pillars.svg @@ -0,0 +1,15 @@ + + AI governance pillars: Policy, Accountability, Lifecycle, Oversight + + Policy + and standards + + Accountability + and decision rights + + Lifecycle + controls + + Oversight + ongoing + diff --git a/site/training/ai-governance/intentory-in-action.gif b/site/training/ai-governance/intentory-in-action.gif new file mode 100644 index 0000000000..5c264681c8 Binary files /dev/null and b/site/training/ai-governance/intentory-in-action.gif differ diff --git a/site/training/ai-governance/inventory-overview.png b/site/training/ai-governance/inventory-overview.png new file mode 100644 index 0000000000..015b3bbadb Binary files /dev/null and b/site/training/ai-governance/inventory-overview.png differ diff --git a/site/training/ai-governance/managing-ai-use-cases.qmd b/site/training/ai-governance/managing-ai-use-cases.qmd index c2bad4d3fa..17bf589319 100644 --- a/site/training/ai-governance/managing-ai-use-cases.qmd +++ b/site/training/ai-governance/managing-ai-use-cases.qmd @@ -15,7 +15,7 @@ format: help: true controls-back-arrows: visible transition: slide - theme: [default, ../assets/slides.scss, ./slides.scss] + theme: [../assets/slides.scss, ../assets/use-cases.scss] slide-number: true chalkboard: false preview-links: auto @@ -56,81 +56,126 @@ This second module is part of a four-part series: {{< include /training/assets/_revealjs-navigation.qmd >}} -# Use case inventories {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +# Use case inventories {background-image="../assets/digitalspace-colours.svg" .title-accent} ## What is a use case inventory? {.center} -A centralized registry of all AI systems and their purposes. +A centralized registry of all AI systems and their purposes, use case inventories help you: -Use case inventories help you: +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Understand where AI is used -- Understand where AI is used -- Track ownership and accountability -- Assess aggregate risk exposure -- Demonstrate governance to regulators +[{{< fa check >}}]{.dark-pink}   Track ownership and accountability + +[{{< fa check >}}]{.dark-pink}   Assess aggregate risk exposure + +[{{< fa check >}}]{.dark-pink}   Demonstrate governance to regulators +::: ## Inventory fields {.center} +:::: {.columns} + +::: {.column width="50%" .pr3 .f2} Configure custom inventory fields for AI governance: +::: {.f4 .mt5} | Field type | Examples | |------------|----------| | Classification | Risk tier, impact level | | Ownership | Use case owner, business sponsor | | Purpose | Intended use, use boundaries | | Status | Lifecycle stage, approval status | +::: + +::: + +::: {.column width="50%" .pl3 .tilt-left .f3} +![Custom field configuration](custom-fields.gif){fig-alt="A short video showing custom field configuration in ValidMind settings" .screenshot} +::: -## Hands-on: Explore the inventory {.center} +:::: -1. Log in to {{< var vm.product >}} -2. Navigate to **Model Inventory** -3. Review the columns and filters available -4. Consider which fields would support your governance needs +## The inventory in action {.center} -# Risk classification {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +:::: {.columns} + +::: {.column width="60%" .pr3 .tilt-right .mt5 .f3} +![The ValidMind inventory](intentory-in-action.gif){fig-alt="A short video showing the ValidMind inventory with AI use cases displaying tier, model owner, and model stage fields" .screenshot} +::: + +::: {.column width="40%" .pl3 .nt2 .f2} +The **Inventory** displays AI use cases with: + +- **Tier** — Risk classification level +- **Model owner** — Accountability assignment +- **Model stage** — Current lifecycle stage +::: + +:::: + +# Risk classification {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Why classify risk? {.center} -Risk classification enables **proportionate governance**. +Risk classification enables **proportionate governance**. Higher-risk AI systems receive: + +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   More rigorous review -Higher-risk AI systems receive: +[{{< fa check >}}]{.dark-pink}   Additional documentation requirements -- More rigorous review -- Additional documentation requirements -- Enhanced monitoring -- Stricter approval gates +[{{< fa check >}}]{.dark-pink}   Enhanced monitoring + +[{{< fa check >}}]{.dark-pink}   Stricter approval gates +::: ## Classification schemes {.center} Align your classification to relevant regulations: +::: {.mt4 .f2} | Framework | Classification levels | |-----------|----------------------| | EU AI Act | Prohibited, high-risk, limited-risk, minimal-risk | | Internal | Critical, high, medium, low | -| Tiered | Tier 1, Tier 2, Tier 3 | +| Tiered | Tier 1, Tier 2, Tier 3, Tier 4 | +::: ## Configuring risk tiers {.center} +:::: {.columns} + +::: {.column width="50%" .pr3 .f2} In {{< var vm.product >}}, you can: -1. Add custom fields for risk classification -2. Configure different workflows per tier -3. Apply documentation templates by tier -4. Generate reports filtered by risk level +1. **Add custom fields** for risk classification +2. **Configure different workflows** per tier +3. **Apply documentation templates** by tier +4. **Generate reports** filtered by risk level +::: -# Impact assessments {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +::: {.column width="50%" .pl3 .mt4} +![Risk overview](analytics-risk-overview.png){fig-alt="A screenshot showing risk tier field configuration in ValidMind" .screenshot .tilt-left} +::: + +:::: + +# Impact assessments {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Purpose of impact assessments {.center} -Impact assessments evaluate potential risks and harms from AI deployment. +Impact assessments evaluate potential risks and harms from AI deployment. They document: + +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Who is affected by the AI system -They document: +[{{< fa check >}}]{.dark-pink}   What decisions the AI influences -- Who is affected by the AI system -- What decisions the AI influences -- Potential for harm or discrimination -- Mitigating controls +[{{< fa check >}}]{.dark-pink}   Potential for harm or discrimination + +[{{< fa check >}}]{.dark-pink}   Mitigating controls +::: ## Impact assessment process {.center} @@ -144,24 +189,25 @@ They document: Use {{< var vm.product >}} to: -- Attach impact assessment documentation -- Track assessment completion status -- Route assessments through approval workflows -- Maintain audit trail of governance decisions +::: {.mt4} +[{{< fa check >}}]{.dark-pink}   Attach impact assessment documentation + +[{{< fa check >}}]{.dark-pink}   Track assessment completion status + +[{{< fa check >}}]{.dark-pink}   Route assessments through approval workflows -# Lifecycle stages {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +[{{< fa check >}}]{.dark-pink}   Maintain audit trail of governance decisions +::: + +# Lifecycle stages {background-image="../assets/digitalspace-colours.svg" .title-accent} ## AI governance lifecycle {.center} -```{mermaid} -graph LR - A[Intake] --> B[Assessment] - B --> C[Approval] - C --> D[Deployment] - D --> E[Monitoring] - E --> F[Review] - F --> B -``` +![](ai-governance-lifecycle.svg){fig-alt="Diagram showing the eight stages of the AI governance lifecycle: Intake (use case owner registers AI system), Assessment (Governance / Risk classify and assess), Documentation (Owner / Risk document model and use case), Validation (Validator test and validate model), Approval (Committee / Compliance sign-off) highlighted in magenta, Deployment (Owner / IT deploy to production), Monitoring (Risk / Ops ongoing oversight), Review (Governance / Audit periodic review), with a dashed arrow from Review back to re-assess or re-approve."} + +::: {.f3 .nt4} +The AI governance lifecycle moves from intake and risk assessment through documentation and validation to a formal approval gate, then deployment, ongoing monitoring, and periodic review — with a feedback loop so systems can be re-assessed and re-approved when needed. +::: ## Managing stage transitions {.center} diff --git a/site/training/ai-governance/risk-classification-workflow.png b/site/training/ai-governance/risk-classification-workflow.png new file mode 100644 index 0000000000..1e5dec02a5 Binary files /dev/null and b/site/training/ai-governance/risk-classification-workflow.png differ diff --git a/site/training/ai-governance/understanding-ai-governance.qmd b/site/training/ai-governance/understanding-ai-governance.qmd index 90ecd998d5..70c62954e1 100644 --- a/site/training/ai-governance/understanding-ai-governance.qmd +++ b/site/training/ai-governance/understanding-ai-governance.qmd @@ -15,7 +15,7 @@ format: help: true controls-back-arrows: visible transition: slide - theme: [default, ../assets/slides.scss, ./slides.scss] + theme: [../assets/slides.scss, ../assets/use-cases.scss] slide-number: true chalkboard: false preview-links: auto @@ -55,58 +55,41 @@ This first module is part of a four-part series: ::: -First, let's make sure you can log in to {{< var vm.product >}}. - {{< include /training/assets/_revealjs-navigation.qmd >}} -## Before you begin {.center} - -::: {.panel-tabset} - -### Registration +# What is AI governance? {background-image="../assets/digitalspace-colours.svg" .title-accent} -To continue, you need to have been [onboarded](ai-governance-register.qmd#register){target="_blank"} onto {{< var validmind.training >}}. +## Defining AI governance {.center} -
+::: {.f2} +AI governance is the **organizational framework** for directing and overseeing how AI is designed, deployed, and used. It sets: -:::: {.tc .mt5 .f2 .embed} -Already logged in and refreshed this module? Click [{{< fa chevron-right >}}]() to continue. +[{{< fa check >}}]{.dark-pink} Policy and standards -::: +[{{< fa check >}}]{.dark-pink} Accountability and decision rights -{{< include /training/common-slides/_log-in.qmd >}} +[{{< fa check >}}]{.dark-pink} Lifecycle controls +[{{< fa check >}}]{.dark-pink} Ongoing oversight ::: -# What is AI governance? {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} - -## Defining AI governance {.center} - -AI governance is the **organizational framework** for directing and overseeing how AI is designed, deployed, and used. - -It sets: - -- Policy and standards -- Accountability and decision rights -- Ethics and compliance requirements -- Lifecycle controls -- Ongoing oversight ## Unit of management {.center} +::: {.f2} In AI governance, the primary unit of management is the **AI system** or **AI use case** — not the individual model. -:::: {.flex .flex-wrap .justify-around} +:::: {.flex .flex-wrap .justify-around .mt4} -::: {.w-40-ns} -**AI governance focuses on:** +::: {.w-40-ns .content-block-teal} +**Focuses on:** - How AI is used - Impact on stakeholders - Organizational accountability ::: -::: {.w-40-ns} +::: {.w-40-ns .content-block-teal} **Applies to:** - Model-based AI @@ -116,38 +99,55 @@ In AI governance, the primary unit of management is the **AI system** or **AI us :::: -# AI governance vs MRM {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +::: + +# AI governance vs MRM {background-image="../assets/digitalspace-colours.svg" .title-accent} ## Parallel use cases {.center} +::: {.f2} AI governance and model risk management (MRM) are **parallel use cases** — not subsets of each other. +::: +::: {.f3 .mt4} | Aspect | AI Governance | MRM | -|--------|---------------|-----| +|----|-------|-------| | Unit of management | AI system / use case | Model | | Objective | Organizational oversight | Technical risk control | | Scope | Broad — ethics, compliance | Narrow — performance, validation | +::: ## Relationship {.center} -AI governance and MRM **overlap** for machine learning models embedded in AI systems. +::: {.f2} +AI governance and MRM may also **overlap** with some shared artifacts: -The overlap may produce shared artifacts: +:::: {.columns} -- Inventories -- Approvals -- Monitoring evidence +::: {.column width="40%"} +- Inventory +- Approval workflows - Issue tracking +- Ongoing monitoring +::: + +::: {.column width="60%" .f3 .tc .nt3} +![](ai-governance-mrm-overlap.svg){width="100%" fig-alt="Diagram: AI Governance and MRM as overlapping circles with shared artifacts in the overlap."} +::: -Organizations may coordinate these use cases or manage them separately. +:::: + +Organizations can coordinate these use cases or manage them separately. **{{< var vm.product >}} supports both approaches**. + +::: -# Key terminology {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +# Key terminology {background-image="../assets/digitalspace-colours.svg" .title-accent} ## AI governance terms {.center} -:::: {.flex .flex-wrap .justify-around} +:::: {.flex .flex-wrap .justify-around .mt4} -::: {.w-40-ns} +::: {.w-40-ns .content-block-teal} **Units of oversight:** - AI system @@ -156,7 +156,7 @@ Organizations may coordinate these use cases or manage them separately. - Automated decision system ::: -::: {.w-40-ns} +::: {.w-40-ns .content-block-teal} **Risk framing:** - AI risk @@ -169,9 +169,9 @@ Organizations may coordinate these use cases or manage them separately. ## Classification and lifecycle {.center} -:::: {.flex .flex-wrap .justify-around} +:::: {.flex .flex-wrap .justify-around .mt4} -::: {.w-40-ns} +::: {.w-40-ns .content-block-teal} **Classification:** - Risk tier @@ -180,7 +180,7 @@ Organizations may coordinate these use cases or manage them separately. - Prohibited / high-risk / limited-risk ::: -::: {.w-40-ns} +::: {.w-40-ns .content-block-teal} **Lifecycle:** - Intake @@ -192,17 +192,32 @@ Organizations may coordinate these use cases or manage them separately. :::: -# Platform orientation {background-color="#083E44" background-image="/training/assets/digitalspace-colours.svg"} +# Platform orientation {background-image="../assets/digitalspace-colours.svg" .title-accent} ## {{< var vm.product >}} for AI governance {.center} +:::: {.columns} + +::: {.column width="50%" .pr3 .f2} {{< var vm.product >}} supports AI governance through: -- **Inventory** — Track AI systems and use cases -- **Custom fields** — Configure risk tiers, impact levels, ownership -- **Workflows** — Intake, approval, and deployment processes -- **Documentation** — Generate governance documentation -- **Dashboards** — Monitor compliance across your portfolio +::: {.f3} +- **Inventory** — Track AI systems use cases, owners, stakeholders, and more +- **Custom fields** — Configure risk tiers, impact levels, and more +- **Workflows** — Intake, approval, and lifecycle processes +- **Documentation** — Run testing and generate documentation +- **Validation** - Identify and track issues +- **Dashboards** — Monitor compliance +::: +::: + +::: {.column width="50%" .pl3 .pt3 .f3} +![The Inventory](inventory-overview.png){fig-alt="A screenshot of the ValidMind Platform showing the main interface" .screenshot .tilt-left} + +- **Document Checker** — Assess model documentation against regulations and internal policies +::: + +:::: ## Next steps {.center} diff --git a/site/training/ai-governance/workflow-approval-step-configuration.png b/site/training/ai-governance/workflow-approval-step-configuration.png new file mode 100644 index 0000000000..345b35008d Binary files /dev/null and b/site/training/ai-governance/workflow-approval-step-configuration.png differ diff --git a/site/training/ai-governance/workflows-periodic-review.png b/site/training/ai-governance/workflows-periodic-review.png new file mode 100644 index 0000000000..0f1e909904 Binary files /dev/null and b/site/training/ai-governance/workflows-periodic-review.png differ diff --git a/site/training/assets/ai-governance-mrm-overlap.svg b/site/training/assets/ai-governance-mrm-overlap.svg new file mode 100644 index 0000000000..c878010085 --- /dev/null +++ b/site/training/assets/ai-governance-mrm-overlap.svg @@ -0,0 +1,17 @@ + + AI governance and MRM as parallel use cases with overlap + + AI Governance + AI system / use case + Organizational oversight + Broad  ethics, compliance + + MRM + Model + Technical risk control + Narrow  performance, validation + + Shared artifacts + inventory, approvals, + monitoring, issues + diff --git a/site/training/assets/slides.scss b/site/training/assets/slides.scss index 0a68f09f7d..9e8fa51958 100644 --- a/site/training/assets/slides.scss +++ b/site/training/assets/slides.scss @@ -182,6 +182,66 @@ section.has-dark-background a:hover { color: #DE257E; } +/* AI governance training: content blocks and table styling (product-agent-demo cues) */ +.reveal .slide .content-block-teal { + background-color: #EDF8FA; + border: 1px solid #60B9C5; + border-radius: 8px; + box-shadow: 0 1px 3px rgba(0, 0, 0, 0.08); + padding: 1rem 1.25rem; +} + +.reveal .slide .content-block-teal strong, +.reveal .slide .content-block-teal b { + color: #083E44; +} + +.reveal .slide .content-block-magenta { + background-color: #FAEDF3; + border: 1px solid #DE257E; + border-radius: 8px; + box-shadow: 0 1px 3px rgba(0, 0, 0, 0.08); + padding: 1rem 1.25rem; +} + +.reveal .slide .content-block-magenta strong, +.reveal .slide .content-block-magenta b { + color: #083E44; +} + +.reveal .slide.title-accent h1, +.reveal .slide.title-accent h2 { + border-left: 5px solid #DE257E; + padding-left: 0.5em; +} + +.reveal .slide table thead th { + background-color: #083E44; + color: white; + border: none; + padding: 0.5em 0.75em; +} + +.reveal .slide table tbody tr:nth-child(odd) { + background-color: #F6FBFD; +} + +.reveal .slide table tbody tr:nth-child(even) { + background-color: #EDF8FA; +} + +.reveal .slide table { + border-collapse: collapse; + border-radius: 8px; + overflow: hidden; + border: none; +} + +.reveal .slide table td, +.reveal .slide table th { + border: none; +} + /* ValidMind URL settings modal and button styles */ /* Settings button (gear icon + text) */ diff --git a/site/training/assets/use-cases.scss b/site/training/assets/use-cases.scss index 19f718dbdf..35bb3cb0b4 100644 --- a/site/training/assets/use-cases.scss +++ b/site/training/assets/use-cases.scss @@ -43,18 +43,20 @@ transform: skew(20deg); } -.reveal #title-slide .subtitle { +.reveal #title-slide .subtitle, +.reveal .title-slide h1 { color: black; -webkit-text-stroke: 5px white; paint-order: stroke fill; position: relative; - display: inline-block; + // display: inline-block; padding: 10px 30px; text-shadow: 0 2px 12px rgba(0,0,0,0.18), 0 1.5px 4px rgba(0,0,0,0.10); font-weight: bold; } -.reveal #title-slide .subtitle::before { +.reveal #title-slide .subtitle::before, +.reveal .title-slide h1::before { content: ""; position: absolute;