Securing Generative AI
Enable practitioners to secure generative AI applications by design: placing controls outside the model, making retrieval respect permissions, limiting what AI tools can do, protecting sensitive data across prompts, outputs and logs, testing before release, and monitoring in production, then documenting it all in a controls plan that can be assured.
Upcoming course. LMS delivery and enrolment are not open. Release dates will be confirmed.
Express interest →
Planned learning outcomes
- Describe the components and trust boundaries of an LLM application and where each security control belongs
- Design retrieval so that answers only draw on content the requesting user is entitled to see
- Apply least privilege and graduated autonomy to the identities and tools an AI application uses
- Protect sensitive data across prompts, retrieved content, outputs, logs and vendor processing
- Plan pre-release adversarial testing as a release gate using defensive test categories
- Specify logging, detections and response actions for a generative AI application in production
- Prepare a generative AI security controls plan with owners, evidence and priorities