Project Elara Generative AI use policy

As per Resolution 2026-09-10.1, Project Elara permits limited use of Generative AI, subject to the following rules:

• We follow Codeberg’s user restrictions on AI use, meaning that vibe coding is not permitted. Generative AI-assisted coding is allowed, as long as code contributions are not fully or overwhelmingly AI-generated.
• If you use Generative AI in the context of your Elara work, please export all chats via an AI Exporter (into markdown format) and upload them to our AI archive
• Do not use AI in a way that violates Project Elara’s Code of Conduct or the Charter, or is otherwise evidently contrary to Project Elara’s mission and values. This includes, but is not limited to, using AI to cheat, plagiarize, fabricate data, make/promote false claims, or harass a member of our community.
• AI-assisted machine translation is allowed as long as the translation is cross-checked by a human translator

Generative AI, if used at all, should be targeted towards uses that will benefit the community and Project Elara as a whole. Some examples of acceptable use of Generative AI include:

• Using it as a learning tool
• Using it for literature search purposes
• Using it for brainstorming, as long as you make sure it provides literature citations and you cite the relevant literature in published research; you should not cite an AI model as it itself is not a reliable source!
• Using it for troubleshooting or code debugging purposes
• Using it to get a better understanding of our codebases and/or to quickly search through our documentation. However, this is to be done through locally-cloned repositories to avoid LLMs querying Codeberg unnecessarily.
• Using it for translating (with human oversight)
• Using it to scrutinize an Elara paper for methodological, experimental, analytical, or other errors/flaws, then making an effort to correct these flaws/errors. This is a particular use of AI that we highly encourage, because it honors our commitment to scientific integrity and producing high-quality research. However, corrections must be made on the basis of peer-reviewed literature sources cited by AI, not by the AI itself, as it is not a reliable source.

Some examples of unacceptable use of AI include:

• Not making a good-faith effort to preserve AI chats in our archive repository
• Vibe coding and/or generating large chunks of code or text (other than for machine-assisted translation). We cannot guarantee the quality of vibe-coded source code. Moreover, it may not be compatible with our public domain licensing, and violates Codeberg’s policies
• Making up fake data/datasets, using it to falsify existing data, or generating fake graphics
• Prompting it to write papers for you, and/or passing off AI-written content/ideas as your work
• Using it to scrape Codeberg for data

We understand that there are always good-faith exceptions to the rule, and will not carry out enforcement actions unless if we believe there is strong justification for doing so. If we have sufficient cause for concern that compels us to take enforcement actions against a member for violations of our AI policy, it will be carried out according to due process as stated in articles 1.4 & 1.5 of the Elara Charter.

Furthermore, this policy is not intended to encourage or endorse the use of Generative AI, which is itself a controversial technology. We believe that the judgement of whether to use it and how much to use it is best left up to the individual judgement of our members, within the confines of our stated AI policy. However, due to the inherent ethical concerns associated with Generative AI, the use of open-source models designed for lower resource consumption and/or self-hosting is strongly encouraged where possible, and members are welcome to not use it at all.