AI is finding vulnerabilities faster than anyone can patch them. Now what?

Anthropic’s announcement of Claude Mythos Preview and Project Glasswing felt like an ominous milestone after a month of historic supply chain attacks. Malware attacks knocked down everyone’s perception of what it means to have a secure software supply chain in 2026. With our approach of building all our artifacts from verifiable source code, Chainguard customers were safe from the Trivy, LiteLLM, Telnyx, and Axios attacks of the last month.

We’re living in a paradox. Patch too fast? Expose your developer secrets to North Korea’s next malware attack. Patch too slow? Fall victim to an exploited CVE. Damned if you do, damned if you don’t.

But then we got Mythos Preview, and the paradox became even starker. The above still holds, plus Anthropic now tells us there’s a tsunami of vulnerabilities that haven’t even been discovered yet, ready to flood all of the software we use. Our CEO, Dan Lorenc, breaks this all down in a video, but this also sums it up pretty well:

How do we solve this riddle?

The vulnerability flood is coming

Mythos Preview identified thousands of zero-day vulnerabilities across every major operating system and web browser, some undetected for decades, including a 27-year-old bug in OpenBSD, a system famous for its security hardening.

As Dan also noted this week, any attacker can now write a working exploit for a published vulnerability in minutes for a few dollars in tokens. Mandiant's latest data confirms that exploitation is, on average, happening before patches are even available. The 90-day disclosure timelines the industry has relied on for years have gone from not useful to totally useless.

For security leaders, the practical implications are immediate. Expect a meaningful uptick in vulnerability disclosures and corresponding patches from your software providers. Your ability to prioritize and deploy those updates quickly will matter more than ever. But you’re still going to face hurdles. Not all of your suppliers will move at the same speed. Smaller or less mature vendors will lag, creating uneven risk across your ecosystem.

The real bottleneck will fall on open source maintainers. AI can find vulnerabilities at machine speed, but the humans who maintain the world's most critical open source projects still operate at human speed, often as volunteers. Anthropic is making Claude Max free for 6 months for open source maintainers to help.

Chainguard can help both OSS maintainers and any organization that wants to build on the safest foundation on the planet. You can start for free with Catalog Starter or Chainguard Libraries.

Meanwhile, the malware side isn't getting easier

With this influx of vulns and patches, you’re going to have to do a lot of updating. But if with that volume, you miss a beat and update from a compromised source, you’ll risk pulling malware into your builds.

The Axios attack tells the story. North Korean threat actors socially engineered a single Axios maintainer, hijacked their npm credentials, and published backdoored versions of a package downloaded over 100 million times per week. The malicious code was live for roughly three hours before it was deleted from npm.

Software supply chain security is a multiplayer game, and sophisticated attackers will always find ways to evade detection. Plus, AI allows them to build more sophisticated schemes to hack open source maintainers. Cooldown policies and malware scanners are useful short-term measures, but they are reactive solutions that require a proactive, preventative solution. When you layer in Mythos, these types of solutions won’t cut it in the long term. This wave of noise to drive patching creates much greater surface area for bad actors to use malware to attack those builds.

What we're hearing from customers

In the days since the Glasswing announcement, we've heard from CISOs who are treating this as a turning point. One Fortune 500 security leader told us they're accelerating Chainguard adoption across every platform and development effort in direct response to Mythos. They were candid: if they were starting from scratch today, rather than two years into the journey, they wouldn't be sleeping. Their concern isn't limited to the discovery of the exploit. They see the open source community under pressure from both directions at once: AI is surfacing vulnerabilities faster than maintainers can respond, while nation-state actors simultaneously target the supply chain itself. Their response has been to put Chainguard front and center with their development teams and mandate adoption.

That reaction captures the shift we're seeing across the industry. Security leaders are making concrete investment decisions based on the reality that the old model of reactive scanning and patching can't keep up.

The attack surface is shifting, and so are we

The barrier to sophisticated cyber activity continues to drop. That makes foundational practices like trusted component registries, signed artifacts with verified provenance, and hardened delivery pipelines more critical than ever. Scanning and patching after the fact isn't enough. You need artifacts that are secure by default.

The faster AI finds vulnerabilities, the more important it becomes to build on a foundation free from known exploits, free from malware, and resistant to new zero-days. But as the industry adopts hardened containers and tightens security around base images, bad actors won't stop. Your security posture is only as strong as your weakest open source dependency, and attackers know this. They'll move to other parts of the software supply chain: libraries, CI/CD actions, agent skills, and other components of the open source ecosystem that haven't received the same level of security focus.

This is exactly why Chainguard is building a complete end-to-end secure software supply chain. We started with containers, added libraries and VMs in 2025, and now offer OS packages, CI/CD actions, and AI agent skills. And, we don’t plan to stop there.

Chainguard is literally built to protect you from the exact damned if you do, damned if you don’t problem. We do daily builds to get containers to zero CVEs. And we build directly from source to keep malware out. Update fast, with safety.

You either build and review all your open source yourself, or you get it from someone who takes care of that for you.

To help organizations get started and get safe quickly, you can get up to five Chainguard container images for free with our Catalog Starter plan, and sign up for console access to get Chainguard Libraries for free until June 30, 2026.