Chainguard has surpassed one billion container build manifests, doubling its output from 500 million in six months as it scales its AI-assisted, self-correcting software supply chain platform.
The milestone also covers more than 3,000 unique container images and 675,000 image versions, reflecting the challenge of keeping cloud-native artifacts current amid constant upstream code, dependency, architecture, and security changes.
A build manifest represents every new verifiable artifact produced by the Chainguard Factory. This includes a newly released language image, a rebuild triggered by a libc patch, an architecture-specific variant, or a regenerated software bill of materials (SBOM) after a dependency update.
Chainguard Hits 1 Billion Container
For heavily used projects such as Python, dozens of supported versions across multiple architectures can require repeated rebuilding as upstream packages evolve.
The company said this continuous rebuild model is designed to address a key gap in conventional vulnerability management. Rather than securing an image only when it is initially pulled, Chainguard aims to ensure that images remain rebuilt, signed, and updated throughout their lifecycle.
Chainguard’s build pipeline starts with Chainguard OS, its Linux distribution designed for cloud-native workloads and rolling software delivery.
Unlike traditional distributions that release on longer cycles, the platform continuously incorporates security, functional, and performance updates from open-source ecosystems.
Artifacts produced by Chainguard Factory are built from source and include SLSA Level 3 provenance, Sigstore signatures, and full SBOMs.
The company relies on declarative, reproducible builds to reduce hidden state and configuration drift, enabling images to be regenerated consistently in response to upstream changes.
However, reproducibility alone was insufficient at catalog scale. Chainguard said its original event-driven Factory architecture increasingly suffered from duplicate work, brittle queues, failed tasks, and operational overhead as dependencies and package relationships expanded.
The company referred to the resulting cycle as a “CVE doom loop,” in which engineers spent significant time resolving build-system decay rather than accelerating patch delivery.
To address those limits, Chainguard introduced Factory 2.0, powered by its open-source DriftlessAF agentic framework.
Instead of reacting independently to each event, the system continuously compares the desired state of the image catalog with its actual state and launches work to close any gap.
That gap may emerge when a CVE is disclosed, a package release appears upstream, a dependency changes, or a new hardening requirement is added. Reconciler bots pull work from a shared queue and use structured, verifiable tooling to build, validate, sign, and publish artifacts.
AI is used for judgment-heavy tasks that are difficult to fully automate with traditional methods, such as assessing newly introduced components, backporting fixes to older releases, and applying lessons from previous successful remediations.
Chainguard said the system is deliberately designed so agents work toward a defined end state, allowing failed tasks to be retried without derailing the overall process.
The push comes as attackers increasingly use AI to identify vulnerable dependencies, analyze attack paths, and accelerate exploit development. In that environment, the time between an upstream patch and the update of a container image becomes a critical security metric.
By expanding its reconciliation-driven build system, Chainguard aims to reduce that exposure window across thousands of interdependent projects.
The company plans to add more reconciler bots, expand the sources feeding its work queue, and move more of its container catalog to the self-healing Factory 2.0 model.
Give your security team the visibility and context to investigate suspicious activity faster and contain threats before business impact grows. Strengthen Your Investigations with ANY.RUN
