Pentesting-Grade Coverage at the Speed of AI

Wednesday, September 2, 2026 | 11:00am ET

AI-assisted development is compressing release cycles and expanding attack surfaces at the same time: more code, more APIs, more integrations, all shipping faster than any human testing cadence can follow.

Attackers have adapted. They've weaponized AI and started looking more for business-logic vulnerabilities that require reasoning about what an application is actually designed to do, probing for them continuously, at machine speed.

Traditional defenses weren't built for this. Point-in-time pentests average 15 days of coverage and cost $15K - $50K+ per engagement, leaving roughly 350 days a year when every new release ships untested at the flaw layer. Point-in-time testing simply can’t cover code that ships every day. This isn’t theoretical: in a recent state-sponsored campaign, AI executed 80–90% of the operation end to end, with almost no human at the keyboard. Average AI breakout time is now measured in minutes, not months.

Join Snyk to see how Evo Continuous Offensive Security (COS) closes that gap: autonomous, reasoning-based security testing at a cost and cadence that match how software is actually built today. And because the hard part of trustworthy AI testing isn’t the model but the system around it, we’ll show why platform context, expert judgment, and independent validation are what turn a capable model into a pen test you can act on.

You’ll also hear directly from Gabriel Brolo Tobar, Staff Security Engineer at Yalo, one of Snyk’s Evo design partners, about the challenges his team is working to solve and why continuous offensive testing matters.