AI speeds up coding but floods open source projects with junk

AI is making programmers faster, but it is also burying the people who run open source projects in poor-quality work. A Linux Foundation survey found maintainers estimate that a quarter of incoming code proposals are low-quality "AI slop." On larger projects, the estimate rises to 31 percent.

What the survey found

Open source software is code shared publicly so anyone can use and improve it. Maintainers are the people who look after a project and decide which changes to accept. A proposed change is called a pull request.

Respondents estimated that AI cut the time spent writing code by 52 percent on average. But 21 percent said highly complex tasks took longer. And 71 percent of maintainers said generative AI has lowered the barriers to contributing. Just over half, 53 percent, disagreed that the technology has improved community trust.

The strain on maintainers

The curl project, a widely used open source tool, announced this week that it will move low and medium severity security flaws to public view. The goal is to protect a security team that AI-assisted reports have almost overwhelmed. Maintainer Daniel Stenberg said all key maintainers feel "strained, overloaded and brought to the brink of what we can muster."

The article notes that Linus Torvalds has warmed to AI tools while staying critical of low-quality submissions.

Rules are patchy

Only 35 percent of projects have a formal policy permitting generative AI use. Another 32 percent have no policy and no plans for one. That leaves expectations unclear for much of the community.

Hilary Carter, who leads research at the Linux Foundation, said projects should plan for a growing number of AI-assisted contributors. She called for clear expectations, putting responsibility for testing and disclosure on contributors, and better tools for maintainers.

What it means for you

Many apps and business systems rely on open source code. When the people reviewing that code are overloaded, careful checking becomes harder. The same applies to any developer or agency you hire. AI can save real time, but unchecked output creates review work for someone else.

• Ask how AI-written code is tested before it ships.
• Ask whether the team discloses when AI helped write it.
• Ask who reviews the code before it goes live.