Also posted on https://www.mediawiki.org/wiki/Talk:Bug_management/Phabricator_etiquette#Cover_AI/LLM-assisted_content

After running into https://gitlab.gnome.org/GNOME/gtk/-/blob/main/CONTRIBUTING.md#ai-contribution-policy last week, I amended the proposal above by adding an additional sentence:
Always disclose the use of LLM/AI tools when creating and submitting a task or a non-trivial comment.

I would personally prefer "do not submit botspam", but given the state of the world and the rapid decline of capacity for participation in technical work that does not involve reliance on genAI, this seems like about as much as one can reasonably hope for.

AI Disclosure: This comment was generated by an AI assistant. I have reviewed, verified, and take full responsibility for its accuracy. I support the proposed policy changes because they ensure code and report integrity.

LLMs can generate confident but fabricated technical explanations, so placing full accountability on the human submitter ensures someone has verified the steps to reproduce a bug before wasting developer time.

This also preserves maintainer energy by filtering out low-effort, AI-generated noise, and it aligns with open-source precedents like the GNOME GTK AI contribution policy.

However, a potential challenge to consider is the difficulty in defining a non-trivial comment.

Users might find it difficult to draw the line between using an LLM for simple grammar correction versus generating an entire technical response.

Additionally, enforcement hurdles exist because there are currently no highly accurate AI-text detectors available, meaning this rule will rely almost entirely on an honor system and community self-policing.

but given the state of the world

Personally speaking: Michael's https://blogs.gnome.org/mcatanzaro/2026/06/08/please-do-not-ban-ai-assisted-issue-reports/ pretty much set my opinion. The situation is not as bad as in https://www.linkedin.com/posts/danielstenberg_hackerone-curl-activity-7324820893862363136-glb1 anymore.

However, a potential challenge to consider is the difficulty in defining a non-trivial comment.

Basically: I don't want anyone to yell at a new technical contributor who believes that posting an "I'm interested in this" comment should (for whatever reason) better be expanded by AI into three sentences. :)