#4849 Create a clear AI Policy

Hello @junegunn,

I'd like to talk about your AI policy, or rather, the current lack of one.

In general, GitHub recommends adding a CONTRIBUTING.md file, so even if fzf doesn't adopt an anti-AI policy, I think that for clarity the position of the fzf project on the use of AI, along with guidelines on disclosure etc, should be explicitly put in a CONTRIBUTING.md file.

At time of writing, I've only been able to detect one usage of AI in the whole of fzf's history, that being commit 290b18d, which was a code review suggestion made by copilot. This suggests that the existing community of contributors and users of fzf are already either anti AI, or don't use it without necessarily being against its use. However, the lack of a clear rules leaves the door open to a vibe coded PR being introduced tomorrow that introduces a major security bug, or deletes the entire contents of users' home directories.

Below are some reasons why using AI can result in poorer code quality and subsequent security incidences:

Poor Code Quality

Vibe coding / agentic workflows result in poorer code quality, and relaxed oversight practices. These effects may be compounded by the common practice of using additional LLM-based tooling to provide code-reviews.

How I Dropped Our Production Database and Now Pay 10% More for AWS
Claude Tested Everything Except the One Thing That Mattered (AI agent refuses to follow explicit instructions to test createPost() in increasingly erratic ways)
Amazon calls engineers for a “deep dive” internal meeting to discuss “GenAI”-related outages
GitClear has released reports in 2024 and 2025 indicating a worsening of key code quality metrics correlating with increased LLM adoption.

Deskilling

There is increasing evidence to show that LLMs negatively impact developers' coding abilities:

Brains show less activity when completing tasks with LLMs compared to completing tasks with search or completing tasks without digital help.
Developers who use early-2025 LLMs reported higher subjective performance, but were measured to have lower objective performance. This gap between subjective and objective performance was considered notable.
In an Anthropic study, learners using LLMs demonstrated lower learning rates on average compared to learners not using LLMs.
A recent study uses the term "cognitive surrender" to describe the way humans tend to offload key critical thinking skills onto LLMs, even when the output is wrong.
A paper entitled "AI Assistance Reduces Persistence and Hurts Independent Performance" from April 2026 by academics from MIT, Oxford, UCLA, and Carnegie Mellon showed alarming evidence that performing a variety of tasks with the help of AI for only 10 minutes causes "inpaired unassisted performance and reduced persistence". The researchers noted that "although AI assistance improves performance in the short-term, people perform significantly worse without AI and are more likely to give up"; they also pointed out that "these findings are particularly concerning because persistence is foundational to skill acquisition and is one of the strongest predictors of long-term learning".

Infosec risks

LLM usage results in massive security holes.

What You Can Do

I'd like to kindly suggest implementing a strict "No AI" policy. (Or, as I've said, at least clarify the position of the fzf project on AI usage.) This can also help with what seems to be larger and larger influxes of low quality drive-by PRs by AI agents.

You can find other projects that have taken steps to say no to AI here: https://noai.starlightnet.work/list.html

For a specific policy, I can recommend the following policies for inspiration:

Thank you and kind regards :pray: -- Y. Meyer-Norwood (@norwd)