Chainguard launches AI-powered security tools as survey reveals widespread unpatched vulnerabilities
The container security firm introduced Harness, which deploys frontier AI models to detect exploitable flaws that traditional scanners overlook, alongside three additional products designed to secure coding agent workflows.
Chainguard unveiled an AI-driven security harness leveraging frontier models, among them Anthropic's restricted Claude Mythos 5.1, to identify and remediate exploitable weaknesses in application code that conventional vulnerability scanners fail to catch. The tool grants developers access to Mythos capabilities without requiring direct model access, though the company has not specified which models perform vulnerability detection for customers outside the restricted program.
According to Chainguard, Harness has uncovered serious flaws in codebases that had previously undergone thorough security assessments, including systems deployed in critical infrastructure. The specifics of the affected projects, the vulnerabilities discovered, and their current remediation status have not been made public.
The Wednesday announcement introduced Harness as part of a four-product rollout that also includes Workspaces, Checks, and expanded Libraries protections. These additions extend Chainguard's platform beyond secured container images and dependency management into agent execution environments and CI verification workflows.
Mythos access without Glasswing
Anthropic has limited Claude Mythos 5.1 access to participants in Project Glasswing, leaving uncertainty about which models will power Harness for customers outside that initiative. The tool will also be offered through Anthropic's Claude Marketplace, enabling teams to apply existing Anthropic spending commitments rather than allocate new budget.
The AI exploit gap
The product announcements coincide with The AI Exploit Gap, a Chainguard-commissioned Wakefield Research survey of 1,300 security leaders across nine markets. The findings revealed that more than half of respondents acknowledged their organizations knowingly leave high or critical vulnerabilities unpatched periodically, with the proportion climbing to 66% among companies with $1 billion or more in annual revenue, compared to 27% at companies with revenue below $250 million.
AI can find and chain flaws in minutes, while fixing them still moves at human speed and competes with product deadlines and uptime
Quincy Castro, Chainguard's chief information security officer
Castro elaborated that patching at accelerated rates remains impractical for most organizations and will increasingly consume resources simply to maintain current security posture.
Isolated microVMs for agents
Each developer and coding agent operates within a hardware-isolated microVM subject to network policies that limit connections to internal systems. Credentials remain external to the agent environment, constraining the reach of a compromised agent or malicious dependency. Signed activity logs track agent actions across multiple autonomous sessions, providing audit visibility.
CI inside the agent loop
Chainguard Checks enables coding agents to execute CI checks during development and obtain signed results within seconds. Agents can run tests, examine failures, and modify code before submitting a pull request, circumventing CI queue backlogs that intensify as vendors including Cognition, Microsoft, and Augment Code deploy coding agents as shared team infrastructure managing multiple concurrent sessions.
Checks employs warm snapshots with pre-installed dependencies to avoid environment reconstruction on each execution, refreshing snapshots in the background when the lockfile is updated. Each workflow step runs in a separate hardware-isolated microVM constructed from Chainguard OS packages, incorporating FIPS-compliant toolchains. Workspaces pauses agent sessions during verification and resumes with structured failure logs available in context.
Runners do not retain long-lived credentials. Instead, a host-side proxy injects short-lived credentials into outbound requests, while declarative egress policies govern network access and can shut down a microVM if it attempts an unauthorized connection. Successful executions generate signed attestations linked to the execution environment, test results, and commit, which merge gates can validate without rerunning tests. Checks is currently available in beta.
Libraries lock down late-breaking bugs
Chainguard Libraries is introducing secure package endpoints for Go and .NET, along with CVE remediation for JavaScript and Go. Support for Java and Python will expand to address newly disclosed vulnerabilities identified through AI-assisted research, including flaws in dependencies that coding agents may have already incorporated into projects.
These enhancements build upon Chainguard Repository, a policy-enforced endpoint for open source dependencies, and its collaboration with Cursor, which allows coding agents to retrieve packages through Chainguard's catalog rather than pulling directly from public registries.