Kunal Pathak

# Data Governance for AI Systems

## Navigating the New Era of AI Regulation

Author - <a href="https://www.kunal-pathak.com" target="_blank" rel="noopener noreferrer">Kunal Pathak</a>

---

# Agenda

1. Overview of the EU AI Act
2. EU's Digital Policy Evolution
3. GDPR + EU AI Act - Practical Impact

Note:

---

# Overview of the EU AI Act

---

## Key Features of EU AI Act

- World's first comprehensive AI regulation
- Main Goal - Safer, ethical AI that respects people's rights
- Balance trust and innovation
- Adopted in 2024 and will be implemented gradually over 2025-2026
- Complements GDPR for AI governance

---

## Risk Categories as per EU AI Act

![](./../assets/023-risk_categories_as_per_eu_ai_act.png)

Note:

```mermaid
flowchart TB
    subgraph "Unacceptable Risk"
        J["Banned AI Practices"]
        K["e.g. Social scoring"]
        L["e.g. Manipulation tools"]
    end

subgraph "High Risk"
        M["Strict Compliance"]
        N["e.g. Hiring tools"]
        O["e.g. Credit scoring"]
    end

subgraph "Limited Risk"
        P["Transparency Required"]
        Q["e.g. Chatbots"]
        R["e.g. Deepfake warnings"]
    end

subgraph "Minimal Risk"
        S["No Requirements"]
        T["e.g. Spam filters"]
        U["e.g. Translation tools"]
    end
```

---

## Compliance for AI/ML Activities

- **Audit Trails**
  - Log AI Decisions and Interactions
  - Maintain Version History
  - Data Quality and Bias Detection
- **Transparency**
  - Enhance Model Interpretability
  - Create User-Friendly Explanations
  - Handle User Inquiries Effectively
- **Compliance**
  - Establish Oversight Team
  - Conduct Regular Audits
  - Version Control for Datasets & Models

---

# EU's Digital Policy Evolution

---

## EU Data Protection Timeline

![](./../assets/024-eu_data_protection_timeline.png)

Note:

```mermaid
timeline
    section Pre-GDPR Era
        1995 : Data Protection Directive
        2002 : ePrivacy Directive
    section GDPR Era
        2016 : GDPR Adopted
        2018 : GDPR Enforcement
    section AI Act Era
        2024 : AI Act Adopted
        2025-2026 : Gradual Implementation
```

---

## Evolution of EU Data Protection

![](./../assets/025-evolution_of_eu_data_protection.png)

Note:

```mermaid
graph TD
    subgraph Era1["Fragmented Era (Pre-GDPR)"]
        A1[Fragmented EU laws]
        A2[Minimal enforcement]
        A3[Limited rights]
        A4[Tech outpaced regulations]
    end

subgraph Era2["Privacy Revolution (GDPR)"]
        B1[Unified standards]
        B2[Individual rights priority]
        B3[Privacy by design]
        B4[Strict penalties]
    end

subgraph Era3["Ethical Frontier (AI Act)"]
        C1[Ethical AI focus]
        C2[Transparency & oversight]
        C3[Quality AI training]
        C4[Innovation vs protection]
    end

Era1 --> |Led to need for unified approach| Era2
 Era2 --> |Evolved toward ethical framework| Era3
```

---

# GDPR + EU AI Act - Practical Impact

---

## Data Governance in AI Age

![](./../assets/026-data_governance_in_ai_age.png)

Note:

```mermaid
flowchart TB

subgraph "Best Practices"
        G["Regular audits"]
        H["Clear explanations"]
        I["Continuous monitoring"]
    end

subgraph "Governance Requirements"
        D["Documentation"]
        E["Risk assessments"]
        F["Human oversight"]
    end

subgraph "Data Quality & Ethics"
        A["Bias detection"]
        B["Fair preprocessing"]
        C["Quality controls"]
    end
```

---

## Navigating the fine line

- **Avoid Literal Compliance**
  - Intrusive cookie banners create “consent fatigue”.
  - Overly strict data deletion risks losing valuable insights.
  - Complex AI disclosures meet compliance but confuse users.
- **Embrace Spirit of Data Governance**
  - User-friendly consent fosters autonomy and trust.
  - Balanced data retention upholds privacy without sacrificing business value.
  - Clear, accessible AI explanations build user confidence.
- **Adopt Ethical Mindset**
  - Put user autonomy at the core of compliance
  - Balance privacy protection with responsible data use
  - Aim for genuine transparency in AI explanations

---

## Key Takeaways

![](./../assets/027-key_takeaways.png)

Note:

```mermaid
graph TD
    A[Key Takeaways] --> B[Integrate compliance early into Data Lifecycle]
    A --> C[Balance innovation with ethics]
    A --> D[Prioritize data quality and transparency]
```

---

# More Details on EU AI Act

- **About the EU AI Act**
	- Download <a href="https://www.kunal-pathak.com/blog/eu-ai-act/" target="_blank" rel="noopener noreferrer">my quick 2-minute introduction</a> to the EU AI Act.
	- Deep dive with <a href="https://github.com/kanad13/EU-AI-Act/blob/main/decks/EU_AI_Act.md" target="_blank" rel="noopener noreferrer">my comprehensive guide to EU AI Act</a>.
- **Data Governance for AI**
	- My presentation about Data Governance in view of <a href="https://kanad13.github.io/EU-AI-Act/decks/data-governance.html" target="_blank" rel="noopener noreferrer">GDPR and EU AI Act</a>.
	- For deeper insights on these topics, visit <a href="https://github.com/kanad13/EU-AI-Act" target="_blank" rel="noopener noreferrer">my GitHub repository</a>.

---

## Thank You

<a href="https://www.kunal-pathak.com" target="_blank" rel="noopener noreferrer">Kunal Pathak</a>