Accelerating auditing and compliance for generative AI on AWS
summary
AWS shows an actionable compliance path: 110 controls across eight domains in AWS Audit Manager that automatically collects CloudTrail, Config and other evidence and exports CSV/PDF audit reports; Bedrock offers model evaluation with human-in-the-loop to score toxicity and accuracy and deploy guardrails for hateful or dangerous outputs; CloudTrail Lake plus CloudWatch and Incident Manager detect data events and trigger escalations; AWS Config custom CFN Guard rules can auto-remediate (example: enable PITR on DynamoDB); Version 2 adds SageMaker support.
AI-written summary · about
chapters (13)
- 0:00Introduction to AI auditing
- 1:21Audit manager framework
- 4:08Regulatory landscape
- 9:13Application architecture
- 14:47Ensuring model accuracy
- 20:54Maintaining model fairness
- 25:10Privacy and governance
- 32:28Resilience in AI systems
- 39:15Responsible AI operations
- 42:57Safety and monitoring
- 45:55Securing AI infrastructure
- 48:02Sustainability in AI
- 52:03Key takeaways and resources
related talks
12mUsing AI/ML to scale governance, risk management, and audits
58mImplementing compliance, assurance, and auditing on
32mContinuously assess controls for risk and compliance
33mCompliance as Code - Lessons Learned From Regulated Organizations
50mOptimizing audits with automation
29mCloud compliance and assurance
37mInnovate Responsibly with Generative AI
20mExplore what’s possible with new AWS generative AI services