• Contents:

AI Standard Compliance Challenges

• Complexity and Opacity
  • Challenging to audit how systems make decisions
• Dynamism and Adaptability
  • AI systems change over time, not static
• Emergent Capabilities
  • Unintended capabilities a system may have
• Algorithm accountability
  • Algorithms should be transparent and explainable
    • Regulations in the EU “Artificial Intelligence Act” and US (several states and cities)
    • Promotes fairness, non-discrimination and human rights
• Unique Risks
  • Privacy violations
  • Misinformation
  • Human Bias: humans who create an AI system can introduce bias
  • Algorithmic Bias: if data is biased (not representative), model can perpetuate bias
    • More reading: https://www.sciencebuddies.org/science-fair-projects/project-ideas/Soc_p030/sociology/bias-in-AI-images
    • Example of GenAI bias: generating a picture of doctors showed lots of men and only one woman

AWS Compliance Certifications

• 💡 Over 140 security standards and compliance certifications
• National Institute of Standards and Technology (NIST)
• European Union Agency for Cybersecurity (ENISA)
• International Organization for Standardization (ISO)
• AWS System and Organization Controls (SOC)
• Health Insurance Portability and Accountability Act (HIPAA)
• General Data Protection Regulation (GDPR)
• Payment Card Industry Data Security Standard (PCI DSS)
• …

Model Cards

• Standardized format for documenting the key details about an ML model
  • Intended use, risk rating of a model, training details and metrics
  • Details about used datasets, their sources, licenses, and any known biases or quality issues in the training data.
  • Helpful to support audit activities
• Examples
  • AWS AI Service Cards
  • SageMaker Model Cards
• Screenshot of a SageMaker Model card