You Cannot Ban AI Code You Cannot Detect

Gentoo tried. NetBSD tried. Debian looked at the 45% false positive rate and walked away.

Two years of debate. Still no policy.

In any other context, this would be considered an organizational failure. An open source project with thousands of active developers, relied on by Ubuntu, Raspberry Pi OS, Mint, and hundreds of other distributions, cannot agree on whether code produced by large language models should be accepted into its repositories. The March 2026 DPL (Debian Project Leader) update explicitly called the topic unresolved. The problems haven't been "painful or widespread enough" to force a decision.

Meanwhile, other projects moved. Gentoo banned AI code outright. NetBSD prohibited it. The EFF required disclosure. The Linux kernel quietly rejected suspicious patches.

I've been working with Linux for over twenty years. I've watched communities tear themselves apart over licensing disputes, init system wars, and corporate influence campaigns. And I think Debian's inability to decide is, paradoxically, the most intelligent response anyone in the open source world has produced on this question.