Home

Defensive AI Security & Governance› Models as Insiders›Trusting the Model the Least: Scaling Containment with Capability

In this series

Models as Insiders

1The Insider Playbook, Transferred: Which Insider-Risk Controls Survive Contact with a ModelResearch
2Diversity Is Not Independence: When a Second Model Actually Reduces RiskResearch
3Assume Subversion: AI Control Protocols and How to Evaluate ThemResearch
4The Emergency Brake: Stopping an Agent Mid-Task Without Making Things WorseResearch
5Disclosure Without a Code Path: Reporting Incidents in Model-Driven SystemsResearch
6Trusting the Model the Least: Scaling Containment with CapabilityResearch

HostExpert

Murali Chillakuru

Press play to begin the conversation.

0:00

18:25

Models as Insiders · 6 / 6

Trusting the Model the Least: Scaling Containment with Capability