|
| 1 | +*This text was [taken from the LLVM project](https://raw.githubusercontent.com/rnk/llvm-project/refs/heads/tool-policy/llvm/docs/AIToolPolicy.md) andi is licensed under a `Creative Commons Attribution 3.0 Unported License`_.* |
| 2 | + |
| 3 | +.. _Creative Commons Attribution 3.0 Unported License: http://creativecommons.org/licenses/by/3.0/ |
| 4 | + |
| 5 | +# Simplicity AI Tool Use Policy |
| 6 | + |
| 7 | +## Policy |
| 8 | + |
| 9 | +rust-simplicity's policy is that contributors can use whatever tools they would like to |
| 10 | +craft their contributions, but there must be a **human in the loop**. |
| 11 | +**Contributors must read and review all LLM-generated code or text before they |
| 12 | +ask other project members to review it.** The contributor is always the author |
| 13 | +and is fully accountable for their contributions. Contributors should be |
| 14 | +sufficiently confident that the contribution is high enough quality that asking |
| 15 | +for a review is a good use of scarce maintainer time, and they should be **able |
| 16 | +to answer questions about their work** during review. |
| 17 | + |
| 18 | +We expect that new contributors will be less confident in their contributions, |
| 19 | +and our guidance to them is to **start with small contributions** that they can |
| 20 | +fully understand to build confidence. We aspire to be a welcoming community |
| 21 | +that helps new contributors grow their expertise, but learning involves taking |
| 22 | +small steps, getting feedback, and iterating. Passing maintainer feedback to an |
| 23 | +LLM doesn't help anyone grow, and does not sustain our community. |
| 24 | + |
| 25 | +Contributors are expected to **be transparent and label contributions that |
| 26 | +contain substantial amounts of tool-generated content**. Our policy on |
| 27 | +labelling is intended to facilitate reviews, and not to track which parts of |
| 28 | +LLVM are generated. Contributors should note tool usage in their pull request |
| 29 | +description, commit message, or wherever authorship is normally indicated for |
| 30 | +the work. For instance, use a commit message trailer like Assisted-by: <name of |
| 31 | +code assistant>. This transparency helps the community develop best practices |
| 32 | +and understand the role of these new tools. |
| 33 | + |
| 34 | +An important implication of this policy is that it bans agents that take action |
| 35 | +in our digital spaces without human approval, such as the GitHub [`@claude` |
| 36 | +agent](https://github.com/claude/). Similarly, automated review tools that |
| 37 | +publish comments without human review are not allowed. However, an opt-in |
| 38 | +review tool that **keeps a human in the loop** is acceptable under this policy. |
| 39 | +As another example, using an LLM to generate documentation, which a contributor |
| 40 | +manually reviews for correctness, edits, and then posts as a PR, is an approved |
| 41 | +use of tools under this policy. |
| 42 | + |
| 43 | +This policy includes, but is not limited to, the following kinds of |
| 44 | +contributions: |
| 45 | + |
| 46 | +- Code, usually in the form of a pull request |
| 47 | +- RFCs or design proposals |
| 48 | +- Issues or security vulnerabilities |
| 49 | +- Comments and feedback on pull requests |
| 50 | + |
| 51 | +## Extractive Contributions |
| 52 | + |
| 53 | +The reason for our "human-in-the-loop" contribution policy is that processing |
| 54 | +patches, PRs, RFCs, and comments to LLVM is not free -- it takes a lot of |
| 55 | +maintainer time and energy to review those contributions! Sending the |
| 56 | +unreviewed output of an LLM to open source project maintainers *extracts* work |
| 57 | +from them in the form of design and code review, so we call this kind of |
| 58 | +contribution an "extractive contribution". |
| 59 | + |
| 60 | +Our **golden rule** is that a contribution should be worth more to the project |
| 61 | +than the time it takes to review it. These ideas are captured by this quote |
| 62 | +from the book [Working in Public][public] by Nadia Eghbal: |
| 63 | + |
| 64 | +[public]: https://press.stripe.com/working-in-public |
| 65 | + |
| 66 | +> \"When attention is being appropriated, producers need to weigh the costs and |
| 67 | +> benefits of the transaction. To assess whether the appropriation of attention |
| 68 | +> is net-positive, it's useful to distinguish between *extractive* and |
| 69 | +> *non-extractive* contributions. Extractive contributions are those where the |
| 70 | +> marginal cost of reviewing and merging that contribution is greater than the |
| 71 | +> marginal benefit to the project's producers. In the case of a code |
| 72 | +> contribution, it might be a pull request that's too complex or unwieldy to |
| 73 | +> review, given the potential upside.\" \-- Nadia Eghbal |
| 74 | +
|
| 75 | +Prior to the advent of LLMs, open source project maintainers would often review |
| 76 | +any and all changes sent to the project simply because posting a change for |
| 77 | +review was a sign of interest from a potential long-term contributor. While new |
| 78 | +tools enable more development, it shifts effort from the implementor to the |
| 79 | +reviewer, and our policy exists to ensure that we value and do not squander |
| 80 | +maintainer time. |
| 81 | + |
| 82 | +Reviewing changes from new contributors is part of growing the next generation |
| 83 | +of contributors and sustaining the project. We want the LLVM project to be |
| 84 | +welcoming and open to aspiring compiler engineers who are willing to invest |
| 85 | +time and effort to learn and grow, because growing our contributor base and |
| 86 | +recruiting new maintainers helps sustain the project over the long term. Being |
| 87 | +open to contributions and [liberally granting commit access][commit-access] |
| 88 | +is a big part of how LLVM has grown and successfully been adopted all across |
| 89 | +the industry. We therefore automatically post a greeting comment to pull |
| 90 | +requests from new contributors and encourage maintainers to spend their time to |
| 91 | +help new contributors learn. |
| 92 | + |
| 93 | +[commit-access]: https://llvm.org/docs/DeveloperPolicy.html#obtaining-commit-access |
| 94 | + |
| 95 | +## Handling Violations |
| 96 | + |
| 97 | +If a maintainer judges that a contribution is *extractive* (i.e. it doesn't |
| 98 | +comply with this policy), they should copy-paste the following response to |
| 99 | +request changes, add the `extractive` label if applicable, and refrain from |
| 100 | +further engagement: |
| 101 | + |
| 102 | + This PR appears to be extractive, and requires additional justification for |
| 103 | + why it is valuable enough to the project for us to review it. Please see |
| 104 | + our developer policy on AI-generated contributions: |
| 105 | + http://llvm.org/docs/AIToolPolicy.html |
| 106 | + |
| 107 | +Other reviewers should use the label to prioritize their review time. |
| 108 | + |
| 109 | +The best ways to make a change less extractive and more valuable are to reduce |
| 110 | +its size or complexity or to increase its usefulness to the community. These |
| 111 | +factors are impossible to weigh objectively, and our project policy leaves this |
| 112 | +determination up to the maintainers of the project, i.e. those who are doing |
| 113 | +the work of sustaining the project. |
| 114 | + |
| 115 | +If a contributor responds but doesn't make their change meaningfully less |
| 116 | +extractive, maintainers should escalate to the relevant moderation or admin |
| 117 | +team for the space (GitHub, Discourse, Discord, etc) to lock the conversation. |
| 118 | + |
| 119 | +## Copyright |
| 120 | + |
| 121 | +Artificial intelligence systems raise many questions around copyright that have |
| 122 | +yet to be answered. Our policy on AI tools is similar to our copyright policy: |
| 123 | +Contributors are responsible for ensuring that they have the right to |
| 124 | +contribute code under the terms of our license, typically meaning that either |
| 125 | +they, their employer, or their collaborators hold the copyright. Using AI tools |
| 126 | +to regenerate copyrighted material does not remove the copyright, and |
| 127 | +contributors are responsible for ensuring that such material does not appear in |
| 128 | +their contributions. Contributions found to violate this policy will be removed |
| 129 | +just like any other offending contribution. |
| 130 | + |
| 131 | +## Examples |
| 132 | + |
| 133 | +Here are some examples of contributions that demonstrate how to apply |
| 134 | +the principles of this policy: |
| 135 | + |
| 136 | +- [This PR][alive-pr] contains a proof from Alive2, which is a strong signal of |
| 137 | + value and correctness. |
| 138 | +- This [generated documentation][gsym-docs] was reviewed for correctness by a |
| 139 | + human before being posted. |
| 140 | + |
| 141 | +[alive-pr]: https://github.com/llvm/llvm-project/pull/142869 |
| 142 | +[gsym-docs]: https://discourse.llvm.org/t/searching-for-gsym-documentation/85185/2 |
| 143 | + |
| 144 | +## References |
| 145 | + |
| 146 | +Our policy was informed by experiences in other communities: |
| 147 | + |
| 148 | +- [Fedora Council Policy Proposal: Policy on AI-Assisted Contributions (fetched |
| 149 | + 2025-10-01)][fedora]: Some of the text above was copied from the Fedora |
| 150 | + project policy proposal, which is licensed under the [Creative Commons |
| 151 | + Attribution 4.0 International License][cca]. This link serves as attribution. |
| 152 | +- [Rust draft policy on burdensome PRs][rust-burdensome] |
| 153 | +- [Seth Larson's post][security-slop] |
| 154 | + on slop security reports in the Python ecosystem |
| 155 | +- The METR paper [Measuring the Impact of Early-2025 AI on Experienced |
| 156 | + Open-Source Developer Productivity][metr-paper]. |
| 157 | +- [QEMU bans use of AI content generators][qemu-ban] |
| 158 | +- [Slop is the new name for unwanted AI-generated content][ai-slop] |
| 159 | + |
| 160 | +[fedora]: https://communityblog.fedoraproject.org/council-policy-proposal-policy-on-ai-assisted-contributions/ |
| 161 | +[cca]: https://creativecommons.org/licenses/by/4.0/ |
| 162 | +[rust-burdensome]: https://github.com/rust-lang/compiler-team/issues/893 |
| 163 | +[security-slop]: https://sethmlarson.dev/slop-security-reports |
| 164 | +[metr-paper]: https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/ |
| 165 | +[qemu-ban]: https://www.qemu.org/docs/master/devel/code-provenance.html#use-of-ai-content-generators |
| 166 | +[ai-slop]: https://simonwillison.net/2024/May/8/slop/ |
0 commit comments