Snyk Evo Drives 60% of New Deals as AI Risks Grow
Snyk’s new security tool now makes up the majority of its new business as companies rush to secure AI agents before they cause major breaches.
Snyk reports that its Evo platform now accounts for 60% of its new customer deals. This shift happens as companies face a growing gap between the speed at which AI generates code and the ability of human teams to review it for security flaws.
The company states that vulnerabilities in its customer base rose by 108% between late 2024 and early 2026. As AI coding agents become more common, they introduce new risks by automatically pulling in unvetted third-party tools and executing actions that previously required human approval.
Independent Validation Becomes Essential
Snyk argues that AI models should not be allowed to grade their own work. The company built Evo to act as an independent validator, checking the actions and tools used by AI agents in real-time. This approach aims to catch issues that traditional security scans miss because they only look at finished code artifacts.
According to GN technics/ai (en-US), this strategy addresses a critical weakness in current AI systems. By separating the generator from the validator, Snyk seeks to prevent autonomous attackers from exploiting low-severity issues that older security models might ignore.
Rapid Deployment in Enterprise Environments
Unlike traditional security software that often takes months to install, Evo is typically live within a single quarter. 76% of new customers have the system running in production before their initial contract period ends. This speed allows banks and other large firms to secure thousands of internal AI tools quickly.
One major US bank is currently using the platform to secure 1,000 internal agent skills for 50,000 developers. The system runs continuous scans on these tools, helping the bank manage risks without disrupting its development workflow.
Balancing Speed and Security Control
The trade-off for this rapid adoption is a reliance on automated decision-making. While Evo reduces the burden on human reviewers, it requires trust in an algorithm that can block or flag actions in real-time. Companies must ensure that these automated checks do not stifle legitimate development while still stopping malicious activity.
As AI agents gain more control over production environments, the need for such independent oversight layers is becoming a standard requirement. The market is shifting from experimental security tools to essential infrastructure that governs how AI systems interact with sensitive data.
