AI Meets Cybersecurity: Building guardrails, fighting threats, and protecting the future | FUSION’25

AI is accelerating everything—including the attack surface. In this FUSION’25 panel, security leaders dig into how to wrap non-deterministic AI with deterministic controls, build governance that doesn’t kill velocity, and use AI to stop sophisticated, high-impact fraud.

00:00–02:48 —Panel intros
02:49–05:00 — Framing the problem: risk vs. reward in AI security
05:01–07:21 — Making AI safe: least-privilege + deterministic guardrails
07:22–08:40 — Context engineering: making AI more deterministic
08:41–10:00 — JumpCloud: using AI to map data sprawl & reduce tech debt
10:01–11:33 — Ōura: AI-assisted threat modeling for lean teams
11:34–13:18 — Telemetry first: logging agent/model/tool calls to manage risk
13:19–15:29 — Abnormal case study: stopping a $40M invoice fraud
15:30–17:29 — Are hyperscalers “safer”? Reality vs. architecture choices
17:30–19:08 — Scale of attacks, observability, and cloud heterogeneity
19:09–21:00 — AI governance committees: lawyers, frameworks, and balance
21:01–22:10 — NIST-aligned governance; why now is a high-risk moment
22:11–22:55 — Selling into governance: sovereignty & data boundaries
22:56–23:34 — What becomes obsolete? Features vs. categories
23:35–24:59 — Partners that matter (CrowdStrike, Wiz, Salesforce, ServiceNow)
25:00–End — The future CISO: convergence with IT & role evolution