ON DEMAND WEBINAR
Duration: 1 hour​

Attackers aren't probing your LLMs in isolation—they're chaining prompt injections into your internal APIs, data sources, and connected actions. A pentester's job is to break what a CISO has built, and that adversarial lens matters more than ever as AI rewires the attack surface faster than defenses can keep up.

The Cobalt 2026 State of Pentesting Report quantifies the gap: AI applications produce high-risk findings at 2.7x the rate of traditional software, and resolution rates sit at a dataset-low 38%. Joe Brinkley, Head of Offensive Security Research, brings the hacker view; Andrew Obadiaru, CISO, brings the program view (including how the top 10% resolve high-risk issues in 10 days while the bottom 10% take 249). Join to learn:

• How attackers exploit LLM apps through chained, multi-step techniques
• What separates the top 10% of security programs from the bottom 10% (it isn't budget)
• How to use offensive testing to validate your AI security investments

SPEAKERS

Andrew Obadiaru
CISO
Cobalt

Joe Brinkley
Head of Offensive Security Research & Community
Cobalt