Securing Generative AI Workloads
summary
A Five-Pillar framework names Infrastructure security, Data protection, Application security, Responsible AI, Regulatory compliance and gives concrete controls: enforce TLS 1.3 and HSM key storage, run quarterly automated posture assessments, classify training/inference data and apply watermarking and provenance tracking, adopt differential privacy or federated learning for personal data, mitigate prompt injection with input sanitization and realtime output scanners, require OAuth 2.0 with PKCE, rate limiting, tamper-evident audit logs, routine red teaming and mapped model/API inventories.
AI-written summary · about
chapters (12)
- 0:00Introduction and Speaker Background
- 0:35The Rise of Generative AI
- 0:48Challenges in Generative AI
- 1:00Introducing the Five-Pillar Framework
- 2:05Infrastructure Security
- 3:05Data Protection Strategies
- 3:51Application Security
- 4:38Responsible AI Practices
- 5:23Regulatory Compliance
- 6:14Practical Implementation Strategies
- 7:28Future Considerations and Emerging Threats
- 9:08Conclusion and Final Thoughts
related talks
20mSecuring generative AI: Privacy and compliance considerations
44mGenerative AI Security — A Practical Guide to Securing Your AI Application
40mSecuring Generative AI: Is it all an Illusion?
45mSecuring and Governing Generative AI: Learnings from Microsoft
38mAccelerate securely: The Generative AI Security Scoping Matrix
39mSecuring AI at Scale: Practical Defenses against Prompt...
19mSecurity controls for generative AI use cases
14mBuilding Transparent and Ethical Generative AI Models in High-Stakes Industries