The Linux Foundation asked maintainers and contributors how generative AI has changed their work. Respondents estimate that AI cuts code-writing time by 52% on average, yet roughly a quarter of incoming pull requests are now plainly weak generated code.
In large projects maintainers put that share even higher, at about 31%. The foundation concludes that the community needs clear AI policies and tooling to cope with the growing stream of changes.
Lower barrier, no extra trust
71% of maintainers say AI has made a first contribution to open source easier. Yet just over half (53%) disagree that the technology has improved trust within communities. The time savings are not universal either: on genuinely complex tasks, 21% said AI made them slower.
Hardly any rules
Only 35% of projects have a formal policy allowing generative AI. Another 32% have no policy and no plans for one, so contributors often have no idea what is expected. The Linux Foundation suggests making contributors responsible for testing and disclosing AI use, while giving maintainers better tools to filter incoming requests.
Not just code
The flood reaches vulnerability reports too. This week the curl project decided to move low and medium severity bugs into public view to relieve a security team nearly buried under AI-assisted reports. Maintainer Daniel Stenberg admitted the key people are working at the limit of what they can manage.
Our take
AI assistants do save time, but review still has to happen, and checking someone else's generated code eats maintainers' hours. If you send patches to the open source libraries your sites and bots depend on, run the tests and be upfront about AI help. Details are in The Register report.