Linux Foundation: AI Speeds Development, but Low-Quality Code Strains Maintainers

A new Linux Foundation study finds that generative AI is delivering big productivity gains for open source software developers, but AI is also creating an influx of low-quality code submissions that need time-consuming human review.

The research found that an estimated 25% of incoming pull requests, which are proposed changes to software projects, are considered low-quality “AI slop.” For larger open source projects, that figure reaches 31%.

The findings point to a growing challenge for open source communities. While AI tools help developers produce code faster, project maintainers face a growing workload as they review submissions to see which changes are suitable for their projects.

The Linux Foundation report, Navigating Open Source Software Development in the Generative AI Era: Challenges and Opportunities for Project Communities, was sponsored by Red Hat, with support from Futurewei, OpenSSF and LF AI & Data.

Pros and Cons of AI in Development

The report found that 59% of maintainers have experienced a rise in contributions since generative AI tools became widely used. Among large projects, 69% reported higher submission volumes.

Worse, these additional contributions don’t always translate into more usable software. The challenge is greater for open source projects that depend on volunteers or small groups of maintainers to evaluate code.

Despite these difficulties, the Linux Foundation research demonstrates the considerable value of AI-assisted development. Respondents estimated that generative AI reduces code-authoring time by an average of 52%. Even tasks outside coding find benefit: time spent on proposal discussions declined by an estimated 19%.

Developers are using these efficiency gains for other work. Some 66% reported using saved time to handle technical debt, while 63% used it to build new features. Another 58% invested the time in learning and improving their skills.

But AI’s benefits depend partly on the complexity of the work. For demanding development tasks, 21% of respondents reported that using generative AI actually required more time.

The lack of formal rules governing AI-generated contributions impacts developers. Only 35% of open source projects have established policies formally permitting generative AI use. A surprisingly large 32% have no AI policy and no plans to develop one.

This governance gap creates doubt about who’s responsible for verifying AI-generated code before submission. Although 61% of contributors report performing detailed reviews and full functional testing, disclosure of AI assistance remains inconsistent.

Trust is another concern. While 71% of maintainers believe AI lowers barriers to open source participation, 53% disagree that the technology has improved trust within their communities.

AI also hasn’t eliminated the need for experienced developers to guide newcomers. Some 46% of maintainers believe generative AI has raised the need for mentorship, compared with 32% who believe it has reduced that requirement.

The Foundation, responding to the research, recommends written contribution policies, clearer disclosure requirements, standardized attribution and more rigorous testing before AI-assisted code is submitted.