Chainguard launches AI-powered flaw hunting, secures coding agents

Chainguard, a startup specializing in secure software supply chains, today rolled out a vulnerability-hunting service powered by Anthropic’s Claude Mythos 5.1 model, alongside two new products for securing AI coding agents and an expansion of its Chainguard Libraries. The announcements came at the company’s Assemble London conference.

The vulnerability-hunting service targets flaws in software dependencies, a persistent pain point for enterprises adopting AI agents. Chainguard’s approach leans on frontier models to automate discovery, a shift from traditional scanning tools that often miss context-specific risks. The two new products—one for agent authentication, the other for runtime monitoring—address gaps that have drawn attention in recent months, including concerns about identity-related breaches tied to AI agents.

The expansion of Chainguard Libraries further reflects the company’s focus on supply-chain hardening, a priority that has seen increased investment across the sector. Whether AI-powered flaw hunting becomes a standard offering—or remains a specialized tool—will depend on how enterprises balance speed and security as agentic workflows proliferate.

Early adopters will likely be enterprises already using its supply-chain tools, but the broader test is whether model-driven approaches can address the risks emerging alongside AI-driven development.

Sources: siliconangle.com