Over the last two years, startups secured substantial institutional backing from leading venture capital firms for unrealistic ideas about how AI can change coding and software development.
Chainguard, an outlier, has raised over $800 million on the contrarian premise. That AI will turn open-source software into a point of failure.
Founded by former Google engineer Dan Lorenc, the cybersecurity startup raised $636 million last year alone. The startup is estimated to be worth $3.5 billion. It rebuilds open-source software with an added layer of security.
Chainguard strips away anything developers don’t need, continuously applies security fixes, digitally signs every release to prove its authenticity and provides a transparent record showing exactly how each piece of software was built.
Today, the company maintains hardened versions of more than 2,200 open-source projects, spanning nearly 400,000 architecture-specific container image versions that enterprises can use as secure replacements for conventional base images.
Rather than relying primarily on vulnerability scanning after software reaches production, Chainguard seeks to eliminate many vulnerabilities before developers build applications on top of it.
Unlike traditional security vendors that charge for scanning or detecting vulnerabilities, Chainguard commercialises access to trusted software itself.
Customers either subscribe to its catalogue of hardened artefacts or license individual images, while its library offering is priced based on the engineering organisation's language ecosystem adoption and size.
The company says its platform now serves leading enterprise customers, including Canva, Snowflake, OpenAI, Hewlett Packard Enterprise and Anduril, reflecting growing demand for secure-by-default open-source software.
According to ZoomInfo, Chainguard’s current revenue stands at nearly $100 million.
Why the Software Supply Chain Is Broken
“Open source is the best form of software out there,” Lorenc tells AIM. “But at the same time, the way people consume open source software is broken in many ways.”
Most developers do not build open-source software directly from source. Instead, they rely on pre-built packages, libraries and container images that pass through multiple intermediaries before becoming part of an application.
“People are consuming open source through a whole messy chain of links and intermediaries between the source and the maintainers before deploying applications,” Lorenc says.
He argues that every additional layer slows the delivery of security updates while creating more opportunities for attackers to compromise software before it reaches developers. AI has only intensified the problem by accelerating both software development and software supply chain attacks.
“It has put the world in this catch-22 situation,” he says. Organisations must either delay software updates and remain exposed to known vulnerabilities or move quickly and risk deploying compromised software.
Chainguard’s answer is to rebuild and maintain open-source software itself.
Instead of waiting for updates to pass through multiple distributors, the company claims to continuously rebuild, verify and distribute trusted versions directly to customers.
“We try to build a safe source for all of open source software,” Lorenc says.
Developers consume those components much like any other open-source dependency, replacing conventional container images with Chainguard-maintained versions without changing their existing development workflows.
Ultimately, Chainguard aims to provide enterprises with a seamless, drop-in replacement for traditional open-source software, allowing them to eliminate typical security compromises without disrupting their established processes.
Recently, it introduced Chainguard Repository, Chainguard Actions, Agent Skills, The Guardener and Chainguard Libraries, extending its secure-by-default approach beyond containers to software packages, AI agents and CI/CD workflows.
Chainguard has also launched Athena, an industry coalition bringing together organisations including Cisco, Cloudflare, Docker, JPMorganChase and PwC to coordinate AI-driven vulnerability remediation.
According to them, Athena has already processed more than 20,000 vulnerability findings, generated over 2,000 patches and supported more than 500 open-source projects.
Why Mythos Changes the Equation
Much of Lorenc’s urgent warning to enterprises consuming open-source software stems from Mythos, Anthropic’s security-focused AI model.
The model has reportedly uncovered thousands of high-severity vulnerabilities across operating systems, browsers, and enterprise software, prompting Anthropic to restrict access to a small group of trusted organisations due to its dual-use potential.
To Lorenc, however, Mythos is significant not because it is simply a better vulnerability scanner, but because it demonstrates what AI is becoming capable of.
If frontier models can discover and combine software flaws at machine speed, the open-source ecosystem will struggle to review, patch and distribute fixes quickly enough.
Enterprises, he believes, will increasingly depend on trusted intermediaries capable of continuously rebuilding, verifying, and distributing secure software.
According to discussions on Hacker News and Reddit, Chainguard enjoys a strong technical reputation among infrastructure and security engineers, who often praise its rapid vulnerability remediation and hardened software images.
At the same time, developers regularly question its pricing, operational trade-offs and the extent to which organisations should depend on a commercial intermediary for open-source software, particularly as AI-driven security tools continue to improve.
While one could argue that models like Mythos reduce the need for companies like Chainguard by enabling organisations to identify vulnerabilities faster than ever before, Lorenc believes the opposite is true.
Historically, many organisations prioritised fixing only high- and critical-severity vulnerabilities, treating lower-risk flaws as acceptable. Lorenc argues that AI has changed that equation.
As models such as Mythos become capable of combining multiple low- and medium-severity vulnerabilities into practical exploits, enterprises can no longer assess vulnerabilities in isolation.
“Patching all of the low and medium vulnerabilities that Mythos can chain together into real working exploits is now something that organisations can’t ignore,” he says.
Lorenc says this shift has made demand for Chainguard's products more urgent, reinforcing the company’s strategy of reducing vulnerabilities before software reaches production rather than relying solely on post-deployment vulnerability management.
Besides the value of Chainguard, he says, extends beyond reducing CVE counts.
Its hardened software images deliberately remove unnecessary components, such as shells, package managers, and excessive privileges, that attackers frequently exploit after gaining an initial foothold.
While those protections are harder to quantify than vulnerability metrics, reducing the overall attack surface becomes increasingly valuable as AI lowers the cost of discovering and combining software flaws.
Lorenc says the goal is to make consuming secure open-source software almost effortless, allowing developers to focus on building applications instead of constantly worrying about software supply chain security.
Your reaction