Hi Alex, Thanks for the pointer:
> Look at lines 276-277: "Do not include AI tool names or model > identifiers in code, comments, or commit messages." I agree that we probably need to reformulate this (depending on the outcome of this discussion), > I wonder if Co-authored-by isn't enough to capture all levels > of human involvement, from shallow to deep, wdyt? What feels irksome to me with this header is that GH treats it very similarly to real (GH) users and shows them as contributors in commit history and various aggregate reports. The ASF guidelines [1] use different language when talking aboiut copyright: AI - "generated", humans - "authored". I think it is important to keep that distinction. Cheers, Dmitri. [1] https://www.apache.org/legal/generative-tooling.html On Mon, Oct 5, 2026 at 12:10 PM Alex Dutra <[email protected]> wrote: > Hi Dmitri, > > Look at lines 276-277: "Do not include AI tool names or model > identifiers in code, comments, or commit messages." > > As for the exact header name, I'm not opposed to Assisted-by, but it's > not a header explicitly listed in the ASF guidelines. I wonder if > Co-authored-by isn't enough to capture all levels of human > involvement, from shallow to deep, wdyt? > > Thanks, > Alex > > On Mon, Oct 5, 2026 at 5:29 PM Dmitri Bourlatchkov <[email protected]> > wrote: > > > > Hi Alex, > > > > I could not quite find the forbidding language in the Polaris AGENTS.md > > file [1]. Do you have a specific line in mind? > > > > I think it is reasonable and advisable to disclose AI involvement in PR > > descriptions or commit messages. > > > > However, I am not sure the "Co-authored-by:" tag is the right device for > > this. It implies authorship, but the ASF guidelines say that the > > contribution should be an "original creation" [2]. > > > > "Generated-by: <Name of Tool> <Version>" implies that the human > > contribution was limited to crafting a particular prompt. I suppose it is > > fine in that particular context. However, if the human involvement was > more > > profound, I'd like to propose something like "Assisted-by: <Name of Tool> > > <Version>". I think it would be valuable for the PR author to further > > describe the level of human involvement in the latter case. > > > > I think that level of detail would be more valuable for reviewers than > long > > degenerate PR descriptions. > > > > Cheers, > > Dmitri. > > > > [1] > > > https://github.com/apache/polaris/blob/5aab91dafecf498c0c94af62dc4c6e0cf9868a96/AGENTS.md > > [2] https://www.apache.org/legal/generative-tooling.html > > > > > > On Mon, Oct 5, 2026 at 10:35 AM Alex Dutra <[email protected]> wrote: > > > > > Hi all, > > > > > > I recently noticed that Apache Polaris and Apache Iceberg have adopted > > > different guidance on disclosure of AI-generated code. > > > > > > While Polaris forbids AI disclosure in commit messages [1], Iceberg > > > encourages this practice [2]. > > > > > > The ASF guidelines themselves seem to have evolved and now recommend > > > AI disclosure in commit messages [3]: > > > > > > > When providing contributions authored using generative AI tooling, a > > > recommended practice is [...] including the phrase “Generated-by: > <Name of > > > Tool> <Version>” or "Co-authored-by: <Name of Tool> <Version>" > > > > > > My question is: should we change Polaris AI-disclosure guidelines to > > > match more closely what the ASF recommends today? > > > > > > Thanks, > > > Alex > > > > > > [1]: > > > > https://github.com/apache/polaris/blob/5aab91dafecf498c0c94af62dc4c6e0cf9868a96/AGENTS.md > > > [2]: > > > > https://github.com/apache/iceberg/blob/2c0dd406fc1269810074006e49c5ded061feedf1/AGENTS.md > > > [3]: https://www.apache.org/legal/generative-tooling.html > > > >
