Linux Foundation study finds 31% of large-project PRs are low-quality AI-generated submissions

The Linux Foundation, in partnership with Red Hat, Futurewei, LF AI & Data Foundation, and OpenSSF, has published a major research report examining how generative AI is reshaping open-source software development practices and community dynamics. The study, based on survey data from maintainers and contributors across the ecosystem, finds that while AI tools deliver measurable productivity improvements—reducing task completion times by 19% to 52%—they simultaneously introduce significant structural pressures on project governance and maintainer capacity.

For projects exceeding 50 contributors, maintainer burnout has overtaken code quality as the primary concern, with 64% of maintainers reporting that AI-assisted contributions have extended review cycles. The research documents that an estimated 25% of all incoming pull requests across projects qualify as low-quality AI-generated submissions, rising to 31% in larger repositories. Despite these friction points, the study also identifies positive trends: 71% of maintainers recognize AI's potential to lower contribution barriers, 61% of contributors conduct thorough pre-submission review of AI-assisted code, and 77% of maintainers actively use AI tools themselves for tasks like vulnerability scanning and logic summarization.

Key findings on AI's impact on open-source maintenance and trust

• Maintainer burden in large projects: 64% of maintainers on projects with 50+ contributors report that AI-assisted contributions have lengthened review cycles, as submissions often appear superficially correct but lack architectural coherence, testing, or system context.
• Prevalence of low-quality AI contributions: An estimated 25% of incoming pull requests across all surveyed projects consist of low-quality AI-generated code characterized as "AI slop," with this figure climbing to 31% in larger, higher-visibility repositories.
• Policy fragmentation across projects: Only 35% of projects have established policies permitting AI-assisted contributions, while 32% have no policy and none planned; 23% intend to implement one, and 10% enforce explicit bans.
• Emerging policy best practices: Among projects with proactive policies, 76% require contributors to assume personal responsibility for functional testing and line-by-line review, and 76% mandate transparent attribution through mechanisms like the Linux kernel's Assisted-by: git commit trailer.
• Contributor diligence and expanded participation: 61% of contributors perform detailed line-by-line reviews of AI-assisted code before submission, while 71% report increased willingness to engage with unfamiliar languages or codebases, and 46% of respondents indicate heightened need for human mentorship rather than diminished need.

Why it matters

This research provides empirical grounding for Linux distribution maintainers, kernel subsystem reviewers, and foundational project stewards who are currently navigating AI-assisted contributions without established norms. The documented productivity dividends—19% to 52% time savings—suggest that projects which implement structured attribution and verification policies, rather than blanket bans, can capture AI's benefits while protecting the human review bandwidth that sustains critical infrastructure. For the broader open-source funding and governance ecosystem, the findings underscore that reinvestment of AI efficiency gains into mentorship and technical debt reduction, rather than pure velocity, may determine long-term project health.