CISA published a guidebook on July 30, 2026 to help federal agencies manage security risks tied to open-source software, including how they should treat open-weight AI models.

 

What happened

CISA released Open Source Software: Security Principles and Practices for federal civilian agencies. The guide explains the benefits of open-source software, including efficiency, cost savings, and security transparency, while also outlining its tradeoffs. It instructs agencies to evaluate the trustworthiness of an open-source project before approving it for use and to track open-source components in their asset management repositories. The guide also covers how agencies should handle patching, including situations where a newly discovered vulnerability has no available patch. It offers direction on how agencies can contribute to open-source projects, produce their own, and secure government reuse rights when contracting for custom software.

 

The backstory

An executive order signed by President Joe Biden, and later amended by President Donald Trump, directed CISA and other agencies to issue open-source security recommendations for federal agencies.

 

Going deeper

The guidance draws a distinction between open-source AI systems and other open-source software. It states that agencies should evaluate open-weight AI models differently because open-source AI licenses do not require the same transparency needed to assess software trustworthiness. CISA also released related guidance covering software bills of materials, isolation of operational technology during a crisis, and updated secure cloud configuration baselines for Google Workspace.

 

What was said

Æva Black, an open-source security expert and former OSS lead at CISA, told CyberScoop the guidance "demonstrates a grounded understanding of the global, diverse, and participatory nature of open source software development, and provides essential guidance for federal agencies to safely use open source during a crucial moment." She said, "Due to recent advances in AI, particularly in large language models capable of finding and exploiting software vulnerabilities, vulnerability management is facing a global crisis." She added that some proprietary software vendors are using this moment "to spread 'fear, uncertainty, and doubt' about open source in order to capture public attention, and, I presume, public money," while stating she believes open-source software "is, and will remain, the safest and most cost-effective means for building large scale public infrastructure."

 

In the know

Open-source software is software that anyone can use, modify, and share. Open-weight AI models release a model's trained parameters publicly, but that openness does not come with the documentation or transparency needed to evaluate how safe or trustworthy the model is.

 

Why it matters

This guidance comes at a time when attackers are using large language models to find and exploit software vulnerabilities faster than agencies can patch them, according to Æva Black. That shift changes things for federal agencies that already rely on open-source components, the risk isn't just unpatched code, it's AI systems that can actively hunt for weaknesses in that code. By singling out open-weight AI models for separate treatment, CISA is acknowledging that the trust model federal agencies use for traditional open-source software doesn't automatically transfer to AI systems, since an AI license alone doesn't guarantee agencies can verify what's actually inside a given model. For agencies deciding whether to deploy open-weight AI on sensitive networks, that distinction could influence procurement and security review decisions.

 

The bottom line

CISA is asking federal agencies to treat open-source software and open-weight AI models as related but distinct risk categories, with AI systems requiring extra scrutiny before deployment on sensitive networks. Agencies that build trustworthiness evaluations and patching processes now will be better positioned as AI-assisted vulnerability discovery continues to change.

 

FAQs

What is open-source software?

Open-source software is software whose source code is publicly available for anyone to use, modify, and share.

 

What does "patching" mean in cybersecurity?

Patching refers to applying updates to software that fix known security vulnerabilities.

 

What is an open-weight AI model?

An open-weight AI model is an AI system whose trained parameters are released publicly.

 

Why would a federal agency use open-source software instead of proprietary software?

Agencies often choose open-source software because it can offer cost savings, efficiency, and the ability to directly inspect the underlying code.

 

What is a software bill of materials (SBOM)?

A software bill of materials is an inventory list of all the components and dependencies that make up a piece of software.