Chainguard Revolutionizes AI Software Security with Factory 2.0

Chainguard Expands Its Scope to AI Software

Chainguard, known for its specialization in open-source software security, is now tackling a new challenge: the protection of open-core software, AI agent capabilities, and GitHub actions. This initiative is part of a broader strategy aimed at securing the entire software ecosystem in response to the rapid evolution of artificial intelligence technologies.

New Horizons for Chainguard

The company recently unveiled its new approach at the Chainguard Assemble 2026 event in Manhattan. Dan Lorenc, co-founder and CEO of Chainguard, illustrated this transition with a practical demonstration. By comparing the use of a hand saw to that of a power saw, he highlighted the increased dangers associated with rapid automation while emphasizing the need to learn how to use these tools safely. According to him, the industry is undergoing a transformation, moving from artisanal methods to automated processes, with AI playing a central role.

Lorenc predicted that within a year, the majority of code would be generated by automated systems. To counter the growing threats posed by AI, he stressed the importance of automating traditionally lengthy security update cycles and designing secure systems from the ground up.

The Innovation of Chainguard Factory 2.0

To realize this vision, Chainguard has introduced Factory 2.0, an evolution of its automatic image-building method for operating systems and applications. This new version has eliminated over 1.5 million vulnerabilities in its clients' production environments, a significant leap from the 270,000 vulnerabilities fixed the previous year. Factory 2.0 utilizes an AI-powered reconciler pipeline to keep systems in an optimal state, whether it involves the absence of known vulnerabilities, specific quality tests, or performance constraints.

Dustin Kirkland, SVP of Engineering at Chainguard, explained that the company has invested in several AI models, such as OpenAI, Claude, and Gemini, to enhance the accuracy of its agents. While the initial agents had a success rate of only 50 to 60%, the failures served as a foundation for the future training of the models.

Innovative Services for Developers

Chainguard has also launched a series of new services aimed at developers, designed to facilitate the creation of safe and efficient software. Among these services, Chainguard OS stands out as a Linux distribution entirely built from source, without reliance on existing distributions like Debian or Fedora. This allows companies to design their own custom Linux distributions, free from bugs.

Kirkland emphasized the importance of this approach by highlighting developer self-service, enabling them to quickly access the necessary software. Chainguard's container catalog remains a central element of its offering, with over 2,200 upstream projects transformed into container images and the maintenance of more than 30,000 OS packages.

A Business Model Tailored to Current Needs

To meet the growing demand, Chainguard has introduced a free tier of its catalog, the ChainGuard Catalog Starter, which allows users to test five free images before committing further. This strategy aims to attract developers looking to experiment before moving to more intensive use.

The company is also turning to Chainguard Commercial Builds, secure images intended for commercial and open-core software, thus moving away from traditional open-source images. Kirkland noted that many clients are looking to integrate Chainguard into their proprietary builds, reflecting a trend towards shared-source or commercial open-source models.

Enhanced Security for CI Systems and AI Skills

Aware of the vulnerabilities in continuous integration (CI) systems, Chainguard has launched two new product families: Chainguard Actions and Chainguard Agent Skills. Lorenc criticized the security model of GitHub actions, highlighting the difficulty of ensuring the reliability of actions available on the market. Chainguard Actions offer a secure alternative, continuously reinforced and tested to ensure the stability of security patches.

Kirkland also mentioned similar challenges in the realm of AI agent skills. Chainguard has encapsulated several hundred of these skills, making them accessible to developers to ensure safe and effective use in software development.