This policy provides general guidance for contributors and maintainers relating to the use of Generative AI in OpenTelemetry projects. This guidance supersedes and extends the policy defined in the Linux Foundation Generative AI Policy.
While we welcome contributions from anyone, maintainers of individual projects may -- at their discretion -- hide or close issues, pull requests, or other contributions that are made totally or in part through generative AI tooling.
Increasingly, we have observed a trend of contributors who are utilizing LLMs and other generative tools to participate in issues and create pull requests. Regurgitating the output of an LLM is unlikely to be particularly helpful, or valuable, to other contributors, maintainers, and end-users for a couple of reasons.
First, time is a very scarce resource for maintainers and approvers. Thoughtful, high-quality code reviews both essential to the success of OpenTelemetry, and require a significant time commitment. There is not enough time to give proper responses to low-effort pull requests without compromising our responsiveness to high-effort pull requests. Second, OpenTelemetry is a complex, fast-moving project -- LLMs will often have stale data in their training sets, and are prone to offering output that is not relevant to the current state of the project. While LLMs can be incredibly powerful coding assistants they are not a substitute for human judgement and knowledge.
With proper usage, Generative AI can be a valuable tool for writing code, documentation, tests, and more. This level of usage requires enough understanding of the project to evaluate the LLM output, and to know when to accept or reject it. Therefore, we ask that contributors do not rely on LLM output as the sole basis for their contributions.
Examples of this include:
- Copying and pasting LLM output into issues or pull requests without any additional context or explanation.
- Reviewing existing pull requests solely via LLMs, or using LLMs to respond to issues without any additional context or explanation.
Q: Can I use LLMs to help me write code, documentation, or tests?
Yes, this policy does not prohibit the use of LLMs to assist in writing code, documentation, or tests. However, we ask that you do not rely on LLM output as the sole basis for your contributions.
Q: Can I use LLMs to help me review pull requests, issues, or understand the code base?
Yes, this is also allowed -- and a good idea! You should use LLMs as a tool to assist in your understanding, but not as a replacement for your own judgement and ability.
Q: How do I know the difference between allowed and disallowed usages of LLMs?
"If you have to ask, you already know the answer." This policy is not a broad ban of LLMs, it is a request that you -- as an individual -- use them in a way that adds value to the project and respects the time of other contributors and maintainers. If you are using LLMs to help you write code, that is fine; You should be clear about this in pull requests and reviews. If you are using LLMs to understand code so that you can participate in issues or reviews, that is also fine -- but you should be clear about this as well. What is not fine is copying and pasting a GitHub issue into an LLM prompt and asking it to write the PR for you, then blindly submitting that response. You must be an active and willing participant in the process of contributing to OpenTelemetry.
Q: Can I close or hide issues or pull requests that are made through LLMs?
Yes, as your discretion you may close or hide issues or pull requests that are made through LLMs. We ask that you provide a clear explanation for why you are doing so, and -- if possible -- provide guidance on how the contributor can improve their contribution.
Q: How do I address contributors who are making consistent, low-effort contributions via LLMs?
If an individual contributor continues to engage in low-effort PRs or issues, and you have exhausted other avenues of communication, please escalate the situation to the OpenTelemetry Governance Committee. Per the Code of Conduct, contributors are expected to help maintain a positive environment, which would include following guidance and published policy.
Q: Can I use LLM or Generative AI tooling to assist in my own work as a maintainer?
In general, you should evaluate the output of LLMs -- regardless of how you use them -- in the same way you'd evaluate the output of a human contributor or non-AI tool. For example, tools like Dosu are being used in certain repositories to aid in code review and issue management. Remember that these tools can make mistakes, and use your best judgement when evaluating their output.