Hi YUan, I guess every AI that I know, if I tell it: Take the mona Lisa and slap an Apache header on it, will do that. That doesn’t make the Mona Lisa Apache 2.0 licensed however 😉
That’s one of the tricky things, especially with new codlings. Where sometimes GPL code is copied from another repo and the header is replaced with the Apache one … which is not quite how this works :-/ Chris Von: Yuan Tian <[email protected]> Datum: Mittwoch, 4. März 2026 um 02:32 An: [email protected] <[email protected]> Betreff: Re: Adding Documentation for AI Coding Agents and Project Standards Hi Chris, Thanks so much for the detailed follow-up and for clearly articulating the core legal concerns around licensing, specifically the risk of inadvertently introducing GPL or other non-Apache licensed code. I completely agree that extreme caution is necessary, especially with new development, and I take your warning about the committer being on the hook very seriously. My original intention with the AGENTS.md documentation was to help mitigate this risk by explicitly enforcing our license headers and project standards on the AI agents. However, I recognize your point about the source of the AI's training data, which is the deeper, more complicated issue. We will proceed with the highest level of care and due diligence. Best regards, ------------------ Yuan Tian On Tue, Mar 3, 2026 at 4:54 PM Christofer Dutz <[email protected]> wrote: > Hi all, > > So, I guess for normal coding tasks it should be ok, however then new > stuff is developed caution needs to be applied. > Because as far as I have seen, AI is also trained on stuff licensed with > proprietary licenses or for example copy-left licenses such as GPL. > AI doesn’t really care about licenses when it comes up with solutions. > However, if it mainly got its information from a GPL licensed codebase, > technically the code it produces based on that, is also GPL licensed. > > I have seen this with experimenting with AI for PLC4X. Here obviously some > information I found with AI I could only find in GPL repositories. > So it’s really quite a gray area. > > That’s also the main concern of Apache. Because if something turns out to > be technically GPL, we’d have to surgically remove the related code later > on. > > So I would say: The person committing is going to be the one on the hook > primarily, but not being careful here can harm the ASF. > > > Chris > > > Von: Yuan Tian <[email protected]> > Datum: Montag, 2. März 2026 um 00:45 > An: [email protected] <[email protected]> > Betreff: Re: Adding Documentation for AI Coding Agents and Project > Standards > > Hi Chris, > > I really appreciate you taking the time to reach out privately and > providing the link to the official Apache policy. It's helpful to know it's > "generally not forbidden." I'll read through the documentation immediately. > > Best, > Yuan > > > On Sun, Mar 1, 2026 at 17:42 Christofer Dutz <[email protected]> > wrote: > > > I took the liberty to reach out privately. > > So it’s generally not forbidden. There’s some information available: > > https://www.apache.org/legal/generative-tooling.html > > > > I’ll read through it aswell 😉 > > > > Chris > > > > > > Von: Yuan Tian <[email protected]> > > Datum: Sonntag, 1. März 2026 um 01:18 > > An: [email protected] <[email protected]> > > Betreff: Re: Adding Documentation for AI Coding Agents and Project > > Standards > > > > Hi Chris, > > > > Thanks for the feedback and for raising the point about the status of AI > at > > Apache. I agree that it’s best to consult with the foundation on this. > > > > Could you please advise on the best channel or process for consulting > with > > Apache about this documentation? > > > > I'll also take a look at speckit—it sounds like a very useful framework > for > > spec-driven development. > > > > Best regards, > > > > Yuan Tian > > > > > > On Sun, Mar 1, 2026 at 00:59 Christofer Dutz <[email protected]> > > wrote: > > > > > Hi Yuan, > > > > > > Please consult with Apache about this … AI is currently still a gray > > space > > > at Apache. Not sure we should formally embrace this just yet. > > > > > > If really embracing agentic coding, I have become a really big fan of > > > speckit. That’s a framework for spec driven development. With human > > > readable documentation of architectural decisions. This was initially > > > intended for documenting decisions for humans. However, it’s become a > > super > > > useful tool for agentic coding and is supported by any agentic > framework > > > that I know. > > > > > > Chris > > > > > > Von: Yuan Tian <[email protected]> > > > Datum: Samstag, 28. Februar 2026 um 05:17 > > > An: dev <[email protected]> > > > Betreff: Adding Documentation for AI Coding Agents and Project > Standards > > > > > > Hi all, > > > > > > I wanted to share a PR that introduces some helpful documentation to > > guide > > > our AI coding models. > > > > > > The goal here is to make sure each model has clear access to our > project > > > guidelines, coding standards, and build procedures. This should help > > reduce > > > friction when contributing and ensure consistency in the codebase. > > > > > > Specifically, we've added `AGENTS.md`, which covers everything from > Java > > > code style and naming conventions to code structure requirements. It > also > > > includes templates for license headers and the necessary commands for > > > verification, building, and running tests. Additionally, `CLAUDE.md` > has > > > been added to include the standard license header and direct readers to > > the > > > comprehensive rules in `AGENTS.md`. > > > > > > You can review the changes here: > > > https://github.com/apache/iotdb/pull/17229 > > > > > > Best regards, > > > -------------------- > > > Yuan Tian > > > > > >
