libvips bans generative-AI contributions over copyright risk

The libvips maintainers merged a commit that replaces the project's previous AI disclosure rules with a flat ban. The new AI Policy in CONTRIBUTING.md reads: "We do not accept contributions containing code or other content made, in whole or in part, using generative AI tools."

The rationale is copyright exposure. "The sources used to train AI models are generally not known, so we cannot determine whether their output may contain material protected by someone else's copyright," the policy states, extending the rule "to this repository and other repositories under the libvips umbrella." The PR, #5203, was opened by kleisauke, who aligned the change with libjpeg-turbo's recent identical ban and described AI-generated code in open source as "legally fraught" at the moment.

Approval was blunt. "This is very strict, and I think I'm happy with that," wrote jcupitt. Contributor dloebl pushed back, describing a workflow of using AI tools to prototype before refining code manually, and asked whether that would still be allowed. Kleisauke said no. The discussion was subsequently locked as too heated. Two researchers we follow in the Who's Who directory circulated the commit link on merge.