Securing AI by design
summary
Cloud security teams learn a concrete framework: discover models and all AWS connections, assess model and application posture (scan models themselves, check permissive configs), and protect at runtime using Prisma AIRS via an inline next‑gen firewall or an API/python package. Claims: 25 prompt‑injection detectors, continuous AI red‑teaming agent, ~90% efficacy on tests. Numbers thrown around: ~50% of orgs build AI, 100,000 apps next year, 2 million Hugging Face models. Demo shows real DLP, URL and agent blocking working in real time.
AI-written summary · about
chapters (9)
- 0:00Introduction and welcome
- 1:06AI adoption trends overview
- 3:08Shadow AI and third-party risks
- 5:59Building secure AI applications
- 9:33Runtime security for AI apps
- 11:21Prisma AIRS framework
- 16:13Deployment options for security
- 17:39Live demo of Prisma AIRS
- 27:32Closing and Q&A invitation
related talks
37mReinforce AI security: Protecting AI applications, models, and data
42mSecure your AI application transformation with Security
18mCommit Secure AI: Embedding Security in Code, Pipelines, and Production
18mSecuring AI models and using AI to maintain compliance
39mSecuring AI at Scale: Practical Defenses against Prompt...
40mEnsuring Data Security in the AI Revolution
14mEmpower AI with Security by Design
35mGenAI Application Security: Not Just Prompt Injection