What AI Security means for you as a developer: Time-focused approach
summary
Demonstrations of adversarial patches and an IR face reflector show how simple physical stickers or reflectors can break camera models (examples: face evasion, a road patch that makes a car swerve, and a sticker that flips a banana into a toaster), while LLMs plus GPU cracking produce real‑time offensive tooling; prediction: most product code becomes machine‑generated in 12–18 months, agentic AI creates autonomous attack surfaces, and the practical rule is trust but verify with provenance, RAG, and a law‑of‑agency constraint.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:31AI security for developers
- 9:45Criminals and AI tech
- 11:58Humans and machines
- 18:06AI system attacks demo
- 19:32AI vision attacks
- 23:42Future of coding with AI
- 28:20Offensive cyber with AI
- 31:50Agentic AI legal needs
- 34:20Advice for devs and managers
- 37:08Q&A
related talks
22mAdversarial AI: Why it Matters Now
57mAI as the New Hacker Frontier
46mA Short Introduction to Hacking Artificial Intelligence
35mHacking AI Systems: How to (Still) Trick Artificial Intelligence
37mFundamentals of AI safety and security
05mThe Asymmetry of AI Security and Physical Hardware Resilience
40mAI As The New Hacker Frontier
42mHow to attack AI systems (and how to defend them) !!!!