We do not accept any contributions that contain content created using Generative AI (referred to as “AI” from here). Examples of products that fit this definition: GitHub Copilot, ChatGPT, Claude, DeepSeek, Llama and Gemini. We distinguish this from machine learning, which this policy does not apply to.
This applies equally to code, writing, graphics, funding applications, or any other contribution. We will close issues/PRs/threads/discussion etc., or otherwise not engage with attempts to contribute in this way.
Limited use of AI may be accepted in the case that: the models are run locally, are openly available, and are only used for limited purposes such as auto-completion, suggestions, or educational use. This usage should be discussed with the Karrot team first.
For external contributors (anyone not in the Karrot Team) who may use AI based tools as part of their workflow, we may accept this use if the final contribution is not substantially AI generated. This should be agreed upon with the Karrot team before contributing.
We ask that contributors honestly declare any use of AI.
The reasons to restrict usage are varied, and some may be personal to team members, but key ones are:
- environmental destruction: impact of data centres on water use, energy use, noise pollution
- skill depletion and subsequent dependence on large capitalist organisations: loss of resilience
- extraction/enclosure of the software commons, and licensing/authorship issues
- poor working conditions for people doing data labelling/collection
- service likely to degrade over time, in the interests of profits: getting people dependent, increase costs, insert ads, etc…
- and more: amplification of social biases, redirection of funding/investment, economic bubble
We also see that it is a changing technology and will continue to evaluate our position as the situation and context evolves.
Consent 2026-09-15
Review 2027-03-16