Breaking Models 🚨 AI Security
Breaking Models 🚨 AI Security (Supply Chain Attacks)
Your AI app runs on parts you didn't build: base models, adapters, datasets, libraries. Agents pull in even more at runtime: tools, MCP servers, prompts, other agents. Each one is a way in.
In this session, we map the AI supply chain end to end: where attackers get in, and what actually stops them.
Agenda
Software vs AI supply chain: the extra steps, and why they widen the attack surface
Threat model walkthrough: data, base models, adapters, conversion and merging, model hubs, deployment
The agent's runtime supply chain: tools, MCP servers, prompt templates, other agents
Real incidents: what broke and how
Defenses and their limits: AIBOMs, signing and provenance, pinning, sandboxed loading, release gates
Open floor: your questions and your own experiences
Built on OWASP GenAI Security Project guidance. Beginners welcome, no ML background needed. No vendor pitches.
📩 Questions, or want to present at a future session? aisecurityinbay@proton.me
📍 WHERE Hacker Dojo, 3200 Patrick Henry Drive, Santa Clara.
🚪 HOW TO GET IN Use the front entrance with the automatic doors and head up to the 2nd floor, Classroom (stairs or elevator).