Blast Radius, Not Blame: Containing Autonomous AI Failures in the Cloud (Nov. 3rd)
Tuesday, November 3rd, 2026: 1:00 PM to 2:00 PM
This session argues that the industry's focus on preventing hallucinations at the model layer is necessary but insufficient - the real control point is runtime.
Virtual
Autonomous AI systems don't fail like traditional software. A hallucination isn't just a wrong answer - it's a confident, plausible, and incorrect output, and when it drives an agent with the authority to act inside your cloud environment, that output becomes a real event: a bad deployment, an unauthorized data pull, a runaway infrastructure change.
Because these systems are probabilistic by design, no amount of pre-production testing, prompt tuning, or policy documentation can guarantee the same failure won't recur tomorrow, on a different input, in a different form.
When cloud infrastructure and AI systems are treated as a single attack surface and secured and observed together in real time, teams can catch a misbehaving agent mid-action, trace exactly what it touched, and cut off the blast radius before one bad decision cascades into an incident.
We'll cover why static guardrails and one-time approvals can't keep pace with probabilistic systems, what runtime security looks like when cloud and AI risk are managed as one problem, and how to shift incident response from "prevent every failure" to "contain any failure, fast."
Hosted by CSA