Open Source Startup Podcast E195: Taking on the New AI Attack Surface With Manifold: Runtime, Skills & Supply Chains
May 26, 2026
Oleks Yaremchuk, co-founder of Manifold Security and AI security engineer, and Neal Swaelens, co-founder and runtime security lead, discuss AI agent runtime and supply chain risks. They cover why runtime detection matters, how Manifest maps skills and plugins, lessons from open projects like LLMGuard, and strategies for building trustworthy AI security products. Short, technical, and forward-looking.
AI Snips
Chapters
Transcript
Episode notes
Founders Met As Neighbors And Built LMGuard
- Neal and Oleks met as neighbors in Berlin and built LMGuard as an open-source entry into AI security.
- They joined Protect AI, were acquired by Palo Alto, then left to found Manifold to focus on runtime agent security and unfinished work.
Agents Shift Security Focus To Runtime Actions
- Agents shift security from safeguarding model responses to securing actions and runtime behavior.
- Endpoints become the battleground because agents pull external skills, plugins, and servers engineers use without centralized visibility.
Discovery Comes Before Detection For Agents
- Discovery of agents and their supply chain is often missing in enterprises, creating the biggest immediate gap.
- Manifold starts customers with discovery, then scans assets so teams can approve or deny skills and plugins.