AI Security, from PoC to Production
summary
This talk covers securing AI agents in production by addressing new attack surfaces introduced by their deployment. It explains concrete protections such as guardrails, AI firewalls, fine-grained access control and continuous monitoring, and presents Battlebox — a generative-AI hardening toolkit from CortAIx integrated into Thales’s pentest solution — with demonstrations of attacks and PoCs.
AI-written summary · about
chapters (15)
- 0:00AI security intro
- 1:27Adversarial attacks
- 5:19LLM attacks
- 7:46Vision attacks
- 10:13Scaling risks
- 13:27AI agent threats
- 17:00Securing AI agents
- 20:09Risk frameworks
- 23:18Attack categories
- 26:11Agent vulnerabilities
- 30:02AI Firewall & Guardrails
- 33:36Risk mitigation
- 36:02Automated security testing
- 40:21Experience & outlook
- 41:02Q&A
related talks
18mCommit Secure AI: Embedding Security in Code, Pipelines, and Production
19mKeynote: Towards Building Safe and Secure Agentic AI
22mFireside Chat - AI Security Vulnerabilities
45mYour AI Agent Has Production Access: Now What? (Jack)
47mInside AI Security with
42mAI for security - friend or foe?
49mThe Intersection of AI and Cybersecurity
42mSecure your AI application transformation with Security