Mitigating Agentic Risk

AI agents are already reading your email, touching your repos, and calling tools on your behalf — usually with more access than the humans who deployed them. This session walks through how agentic systems actually fail in the wild: prompt injection that turns a helpful assistant into an insider threat, tool calls nobody logged, and credentials scoped for convenience instead of least privilege. We'll cover what to instrument, what to block, and what a realistic containment strategy looks like before your org ships its first agent to production.