CONTRIBUTION.md: some sentences about LLM by kfessel · Pull Request #22004 · RIOT-OS/RIOT

kfessel · 2026-01-19T20:56:46Z

Contribution description

Testing procedure

testing the declaration

Declaration of AI-Toolusage 1 :

This PR was written using LLM assistance:
- the LLM was tasked to produce whole file(s),
- the LLM was tasked to produce function(s),
- the LLM was tasked to produce codeblocks or snippets,
- the LLM was used to check the code,
- the LLM was used to produce documentation,
- the LLM was used to spellcheck.
No LLM ("AI") was used in any way.

Tools that were used are:
none

Declaration of AI-Toolusage 2 :

A LLM was used to produce whole file(s).
A LLM was used to produce function(s).
A LLM was used to produce codeblocks or snippets.
A LLM was used to check the code.
A LLM was used to produce documentation.
A LLM was used to spellcheck.
No LLM ("AI") was used in any way.

Tools that were used are:
none

Declaration of AI-Toolusage 3:

Note

Please delete the sentences that are untrue

A LLM was used to produce whole file(s).
A LLM was used to produce function(s).
A LLM was used to produce codeblocks or snippets.
A LLM was used to check the code.
A LLM was used to produce documentation.
A LLM was used to spellcheck.
No LLM ("AI") was used in any way.

Tools that were used are:
none

Declaration of AI-Toolusage short:

Note

Please delete the sentences that are untrue

A LLM was used to produce whole file(s), function(s), codeblocks or snippets.
A LLM was used to produce documentation.
A LLM was used to check the code.
No LLM ("AI") was used in any way but spellcheck

Tools that were used are:
none

Declaration of AI-Toolusage supershort:

Note

Please remove the parts of sentences that are untrue

A LLM was used to produce whole file(s), function(s), codeblocks, snippets, documentation, check the code and spelling.

AI-Tools / LLMs that were used are:

none

Declaration of Generative AI-Tool-usage weekly short:

Note

Please delete the sentences and parts of them that are untrue.
Please add information (e.g., name tools or technologies) if needed.

A LLM agent generated this entire PR either on behave of a human or selftriggered.
A LLM was used to produce whole file(s), function(s), codeblocks or snippets.
A LLM was used to produce documentation.
A LLM was used to check the code.
another kind of Generative AI-Tool was used to do something.
No Generative AI-Tool was used in any way but spellcheck

Tools that were used are:
none

Issues/PRs references

meeting-notes

AnnsAnns

I'm sorry if this looks like a lot but we agreed to look at the details within the PR so I did 😛 Thank you for the PR though.

CONTRIBUTING.md

AnnsAnns · 2026-01-20T10:06:20Z

.github/PULL_REQUEST_TEMPLATE.md

+Declaration of AI-Toolusage:
+[] This PR was written using LLM assistance,
+[] the LLM was tasked to produce whole file(s),
+[] the LLM was tasked to produce function(s),
+[] the LLM was tasked to produce codeblocks or snippets,
+[] the LLM was used to check the code,
+[] the LLM was used to produce documentation,
+[] the LLM was used to spellcheck.
+Tools that where used are: 


I'd personally argue that this is a bit too much, the differentiators don't feel significant enough to have 7 different options

If we only would have "this PR was written using LLM assistance" and assume not knowlegebase based AI was used other wise LLM spell checker would carry the same sentence as "I like a riot driver for IC128 can you produce a PR for me"

What's the consequence of selecting one of these options though?
Do we collect statistics, is there some action associated with one of those 'levels'?

AnnsAnns · 2026-01-20T10:14:02Z

CONTRIBUTING.md

+Large language models are known to sometimes reproduce large amounts of code,
+that was feed into their database as such it might produce a copy and 
+the pervious sentence must be applied.


I don't really understand this sentence

that was feed into their database as such it might produce a copy and the pervious sentence must be applied.

llm ... reproduce large amounts of code that was feed into their database.

i think the , is wrong

CONTRIBUTING.md

AnnsAnns · 2026-01-20T10:16:30Z

CONTRIBUTING.md

+
+
+If a PR was written using tooling that is able to produce large of code from a knowlegebase 
+(e.g.: large language model assisted tools like copilot, devin or cursor or LLM like GPT, Ollama)


Isn't Ollama a tool to run LLMs itself?

i wanted the example list to convey that it is about full tools and also about basic tools like a chat or any other LLM run locally ( the reason for mentioning ollama) or remote (gpt)

(running local does not change the rights issue, having a tool integrated in your editor does not change the issue)
as well doesn't the opposite remote or less integration

I'd restructure the sentence to something like:

If a PR was written using tooling that is able to produce code without attribution (e.g.: LLM tools like Copilot, Devin, Cursor, Claude, ChatGPT or Ollama)

CONTRIBUTING.md

Co-authored-by: Ann🐸 <git@annsann.eu>

kfessel · 2026-01-20T12:23:38Z

.github/PULL_REQUEST_TEMPLATE.md

 -->
+
+Declaration of AI-Toolusage:
+[] This PR was written using LLM assistance,


This PR was written using LLM assistance to produce whole file(s), function(s), codeblocks or snippets, to check the code, produce documentation, spellcheck.

.github/PULL_REQUEST_TEMPLATE.md

CONTRIBUTING.md

kfessel · 2026-01-20T12:34:00Z

I'm sorry if this looks like a lot but we agreed to look at the details within the PR so I did 😛 Thank you for the PR though.

Than you for the review this PR needs input (-> published as draft)

elenaf9 · 2026-01-21T08:38:22Z

.github/PULL_REQUEST_TEMPLATE.md

+This PR was written using LLM assistance:
+- [ ] the LLM was tasked to produce whole file(s),
+- [ ] the LLM was tasked to produce function(s),
+- [ ] the LLM was tasked to produce codeblocks or snippets,
+- [ ] the LLM was used to check the code,
+- [ ] the LLM was used to produce documentation,
+- [ ] the LLM was used to spellcheck.- [ ] This PR was written using LLM assistance,
+- [ ] the LLM was tasked to produce whole file(s),
+- [ ] the LLM was tasked to produce function(s),
+- [ ] the LLM was tasked to produce codeblocks or snippets,
+- [ ] the LLM was used to check the code,
+- [ ] the LLM was used to produce documentation,
+- [ ] the LLM was used to spellcheck.


Suggested change

This PR was written using LLM assistance:

- [ ] the LLM was tasked to produce whole file(s),

- [ ] the LLM was tasked to produce function(s),

- [ ] the LLM was tasked to produce codeblocks or snippets,

- [ ] the LLM was used to check the code,

- [ ] the LLM was used to produce documentation,

- [ ] the LLM was used to spellcheck.- [ ] This PR was written using LLM assistance,

- [ ] the LLM was tasked to produce whole file(s),

- [ ] the LLM was tasked to produce function(s),

- [ ] the LLM was tasked to produce codeblocks or snippets,

- [ ] the LLM was used to check the code,

- [ ] the LLM was used to produce documentation,

- [ ] the LLM was used to spellcheck.

This PR was written using LLM assistance:

- [ ] the LLM was tasked to produce whole file(s),

- [ ] the LLM was tasked to produce function(s),

- [ ] the LLM was tasked to produce codeblocks or snippets,

- [ ] the LLM was used to check the code,

- [ ] the LLM was used to produce documentation,

- [ ] the LLM was used to spellcheck.

Does it really matter if a regular spellchecker or an LLM was used to spellcheck? If not then I don't think the spellcheck checkbox is needed.

Does it really matter if a regular spellchecker or an LLM was used to spellcheck? If not then I don't think the spellcheck checkbox is needed.

Sorry, just saw #22004 (comment).

Co-authored-by: Elena Frank <elena.frank@proton.me>

.github/PULL_REQUEST_TEMPLATE.md

CONTRIBUTING.md

Co-authored-by: Michael Richardson <mcr@sandelman.ca>

CONTRIBUTING.md

.github/PULL_REQUEST_TEMPLATE.md

mguetschow · 2026-01-22T09:59:22Z

.github/PULL_REQUEST_TEMPLATE.md

+- This PR was written using LLM assistance:
+  - [ ] the LLM was tasked to produce whole file(s),
+  - [ ] the LLM was tasked to produce function(s),
+  - [ ] the LLM was tasked to produce codeblocks or snippets,
+  - [ ] the LLM was used to check the code,
+  - [ ] the LLM was used to produce documentation,
+  - [ ] the LLM was used to spellcheck.
+- No LLM ("AI") was used in any way.


Suggested change

- This PR was written using LLM assistance:

- [ ] the LLM was tasked to produce whole file(s),

- [ ] the LLM was tasked to produce function(s),

- [ ] the LLM was tasked to produce codeblocks or snippets,

- [ ] the LLM was used to check the code,

- [ ] the LLM was used to produce documentation,

- [ ] the LLM was used to spellcheck.

- No LLM ("AI") was used in any way.

- [ ] An LLM was tasked to produce whole file(s),

- [ ] An LLM was tasked to produce function(s),

- [ ] An LLM was tasked to produce codeblocks or snippets,

- [ ] An LLM was used to check the code,

- [ ] An LLM was used to produce documentation,

- [ ] An LLM was used to spellcheck.

- [ ] No LLM ("AI") was used in any way.

how about this? I'd argue it is easier to check a checkbox than to remove one or several lines, I already see a lot of PRs with both bullet points still present.

On the other hand, this might also clutter the PR description of every PR, even those that declare "no AI usage" a bit too much.

Btw, why not generalizing to "AI" instead of saying LLM?

To reduce the clutter, we could maybe merge the first three points, and people can remove parts of the sentence if it does not apply.

And I agree with @elenaf9 that spellchecking may not be worth declaring, so we could remove that and maybe change the last line to allow spellchecking.

I'd also merge the first three points + remove "check code" & "spellcheck"

I already see a lot of PRs with both bullet points still present.

I think this is a feature indicating that the pr author did not yet check.

If we join the frist three how do we handle "Threshold of originality" (De: Schöpfungshöhe) and concepts like fair use.

Or would we just assume not enough originality if "A LLM was used to produce whole file(s), function(s), codeblocks or snippets."

Btw, why not generalizing to "AI" instead of saying LLM?

AI is a meaningless marketing term at this point.
LLM is a very specific technical thing.

To reduce the clutter, we could maybe merge the first three points, and people can remove parts of the sentence if it does not apply.

Or would we just assume not enough originality if "A LLM was used to produce whole file(s), function(s), codeblocks or snippets."

IMHO we should still differentiate between generating whole files vs function/ codeblocks / snippets.
The former definitely violates originality and could be reason to reject a PR, the latter not necessarily?

Edit: my logic behind this differentiation is that usually (independently of LLM usage) if I just add function or code snipped to a file I wouldn't add myself to the authors list in the header, but if I write a whole file I'd have the copy-right.

AnnsAnns

Second review (Some comments from the first review are still open)

AnnsAnns · 2026-01-22T09:49:39Z

.github/PULL_REQUEST_TEMPLATE.md

 -->
+
+### Declaration of AI-Toolusage:
+ <!-- You may delete the either sentence in case. --> 


Suggested change





Whitespace (I also don't really understand the sentence itself)

Is a hint to the PR author that they do not need to declare LLM use if they didn't.

following it are two senteces

'This PR was written using LLM assistance ... ' ( the sentence goes on listing option of llm use)
the other

'No LLM ("AI") was used in any way.'

only one of then is true -> the other one may be deleted when creating the PR

.github/PULL_REQUEST_TEMPLATE.md

CONTRIBUTING.md

AnnsAnns · 2026-01-22T09:56:58Z

CONTRIBUTING.md


+## Copyright and AI
+
+RIOT itself applies the LGPL license, see [LICENSE.md], to most of its code exclusively,


Suggested change

RIOT itself applies the LGPL license, see [LICENSE.md], to most of its code exclusively,

RIOT is licensed under the LGPL license, see [LICENSE.md], unless stated otherwise.

AnnsAnns · 2026-01-22T09:58:37Z

CONTRIBUTING.md

+## Copyright and AI
+
+RIOT itself applies the LGPL license, see [LICENSE.md], to most of its code exclusively,
+authors of PRs are assumed to do so as well (not necessary exclusive).


Suggested change

authors of PRs are assumed to do so as well (not necessary exclusive).

I'd honestly just remove this, the same statement is covered in the next sentence.

AnnsAnns · 2026-01-22T10:04:40Z

CONTRIBUTING.md

+Large language models (LLM, "AI") are known to sometimes reproduce large amounts of code
+that was feed into their database as such it might produce a copy and
+the previous sentence must be applied.
+When reworking a PR and copying its code to a new one the author of that code still is the original author
+and their rights to the code must be respected such as naming them and keeping the license.


Suggested change

Large language models (LLM, "AI") are known to sometimes reproduce large amounts of code

that was feed into their database as such it might produce a copy and

the previous sentence must be applied.

When reworking a PR and copying its code to a new one the author of that code still is the original author

and their rights to the code must be respected such as naming them and keeping the license.

Large language models (LLM, "AI") are known to sometimes

reproduce large amounts of code that was fed into their database.

As such it might produce copies of copyrighted works

which do not comply with the license of RIOT.

When reworking a PR or copying code to a new one

the author of that code continues to be the original author

and their rights to the code must be respected.

This includes naming them and respecting their license.

Suggested change

Large language models (LLM, "AI") are known to sometimes reproduce large amounts of code

that was feed into their database as such it might produce a copy and

the previous sentence must be applied.

When reworking a PR and copying its code to a new one the author of that code still is the original author

and their rights to the code must be respected such as naming them and keeping the license.

Large language models (LLM, "AI") are known to sometimes

reproduce large amounts of code that was fed into their database.

As such it might produce copies of works that are

[original](https://en.wikipedia.org/wiki/Threshold_of_originality)

and already licensed incompatible to RIOT or are not licensed at all.

When reworking a PR or copying code to a new one

the author of that code continues to be the original author

and their rights to the code must be respected.

This includes naming them and respecting their license.

Read work as work that passes the "Threshold of originality".

It is about compiling with the license of the thing that was copied not about RIOTs license. For RIOT the work must be either co- or re- license-able to LGPL or keep its license and be link-able (vendor files) so the original license must allow for that.
There also is no not protected work if original (in germany, in the US its possible to declare your own work public domain)
If a work has no license it is unlicensed -> its use is unlicensed. A work does not become copyrighted by declaring copyright but by being created (even in the US) if a work is created by an unknown author with no copyright declared they still got the copyright (until sold) and the moral right (until death (in Europe for N years) (or US where they can be sold or public domain))

Adding a copyright note helps your case in court (Notes are assumed to be true unless proven) but the authors rights are there even without (just need to prove) and without a license a original work is protected (from many things).

CONTRIBUTING.md

Co-authored-by: mguetschow <mikolai.guetschow@tu-dresden.de> Co-authored-by: Ann🐸 <git@annsann.eu>

CONTRIBUTING.md

Co-authored-by: Martine Lenders <martine.lenders@tu-dresden.de>

benpicco · 2026-01-26T22:55:44Z

CONTRIBUTING.md

+authors of PRs are assumed to do so as well (not necessary exclusive).
+When a PR is published it must also respect all authorship-rights, copyrights and
+licenses of code it uses. If code is copied the original author usually must be named and
+the original license must be kept, unless the original license states differently.


I think we should tone it down a bit for the common pattern of copying an existing board / driver /example and starting from there.

We aren't interested in the pedigree of the boilerplate, if it ends up being a complete rewrite the author of the skeleton is just a distraction.

I've seen this a couple of times in the past and always suggested to drop the old @author when they obviously had nothing to do with the new module (and might get asked to support it).

CONTRIBUTING.md

Co-authored-by: Martine Lenders <martine.lenders@tu-dresden.de>

.github/PULL_REQUEST_TEMPLATE.md

Integrate weekly hints

doc: some sentences about LLM to CONTRIBUTION.md

924a3c0

github-actions bot added the Area: doc Area: Documentation label Jan 19, 2026

kfessel changed the title ~~doc: some sentences about LLM to CONTRIBUTION.md~~ CONTRIBUTION.md: some sentences about LLM Jan 19, 2026

crasbe added the Type: enhancement The issue suggests enhanceable parts / The PR enhances parts of the codebase / documentation label Jan 19, 2026

crasbe requested a review from AnnsAnns January 19, 2026 21:23

AnnsAnns requested changes Jan 20, 2026

View reviewed changes

f- Apply suggestions from code review

95b8d7c

Co-authored-by: Ann🐸 <git@annsann.eu>

kfessel commented Jan 20, 2026

View reviewed changes

.github/PULL_REQUEST_TEMPLATE.md Outdated Show resolved Hide resolved

kfessel commented Jan 20, 2026

View reviewed changes

CONTRIBUTING.md Outdated Show resolved Hide resolved

kfessel added 2 commits January 20, 2026 13:31

f Apply suggestions from code review

728806d

f- list rendring Apply suggestions from code review

181e8d8

f Update PULL_REQUEST_TEMPLATE.md

62887cd

elenaf9 reviewed Jan 21, 2026

View reviewed changes

f Apply suggestion from @elenaf9

9135733

Co-authored-by: Elena Frank <elena.frank@proton.me>

mcr reviewed Jan 21, 2026

View reviewed changes

.github/PULL_REQUEST_TEMPLATE.md Outdated Show resolved Hide resolved

mcr reviewed Jan 21, 2026

View reviewed changes

CONTRIBUTING.md Outdated Show resolved Hide resolved

kfessel and others added 3 commits January 21, 2026 17:53

f Apply suggestions from code review

55a6f0f

Co-authored-by: Michael Richardson <mcr@sandelman.ca>

f Apply suggestions from code review

fcc133d

f Update PULL_REQUEST_TEMPLATE.md

22c4474

miri64 reviewed Jan 21, 2026

View reviewed changes

CONTRIBUTING.md Outdated Show resolved Hide resolved

mguetschow reviewed Jan 22, 2026

View reviewed changes

AnnsAnns requested changes Jan 22, 2026

View reviewed changes

f Apply suggestions from code review

998f963

Co-authored-by: mguetschow <mikolai.guetschow@tu-dresden.de> Co-authored-by: Ann🐸 <git@annsann.eu>

miri64 reviewed Jan 22, 2026

View reviewed changes

CONTRIBUTING.md Outdated Show resolved Hide resolved

Update CONTRIBUTING.md

7c11752

Co-authored-by: Martine Lenders <martine.lenders@tu-dresden.de>

benpicco reviewed Jan 26, 2026

View reviewed changes

miri64 reviewed Jan 27, 2026

View reviewed changes

CONTRIBUTING.md Outdated Show resolved Hide resolved

Update CONTRIBUTING.md

1be3d32

Co-authored-by: Martine Lenders <martine.lenders@tu-dresden.de>

basilfx mentioned this pull request Feb 1, 2026

drivers/modbus: add initial support (v3) #22032

Open

benpicco reviewed Feb 13, 2026

View reviewed changes

.github/PULL_REQUEST_TEMPLATE.md Outdated Show resolved Hide resolved

kfessel added 5 commits February 13, 2026 16:29

Update CONTRIBUTING.md

7943003

Integrate weekly hints

f

10291a0

f

24a4ce2

f ben

6f6e8a0

f contrib

5b3bbfc



		If a PR was written using tooling that is able to produce large of code from a knowlegebase
		(e.g.: large language model assisted tools like copilot, devin or cursor or LLM like GPT, Ollama)

	<!-- You may delete the either sentence in case. -->
	<!-- You may delete the either sentence in case. -->


		## Copyright and AI

		RIOT itself applies the LGPL license, see [LICENSE.md], to most of its code exclusively,

	RIOT itself applies the LGPL license, see [LICENSE.md], to most of its code exclusively,
	RIOT is licensed under the LGPL license, see [LICENSE.md], unless stated otherwise.

-Large language models (LLM, "AI") are known to sometimes reproduce large amounts of code
-that was feed into their database as such it might produce a copy and
-the previous sentence must be applied.
-When reworking a PR and copying its code to a new one the author of that code still is the original author
-and their rights to the code must be respected such as naming them and keeping the license.
+Large language models (LLM, "AI") are known to sometimes
+reproduce large amounts of code that was fed into their database.
+As such it might produce copies of copyrighted works
+which do not comply with the license of RIOT.
+When reworking a PR or copying code to a new one
+the author of that code continues to be the original author
+and their rights to the code must be respected.
+This includes naming them and respecting their license.

Conversation

kfessel commented Jan 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Contribution description

Testing procedure

testing the declaration

Declaration of AI-Toolusage 1 :

Declaration of AI-Toolusage 2 :

Declaration of AI-Toolusage 3:

Declaration of AI-Toolusage short:

Declaration of AI-Toolusage supershort:

Declaration of Generative AI-Tool-usage weekly short:

Issues/PRs references

Uh oh!

AnnsAnns left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

AnnsAnns Jan 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

kfessel commented Jan 20, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

elenaf9 Jan 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

AnnsAnns left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

kfessel commented Jan 19, 2026 •

edited

Loading

AnnsAnns Jan 20, 2026 •

edited

Loading

elenaf9 Jan 23, 2026 •

edited

Loading

kfessel Jan 22, 2026 •

edited

Loading

benpicco Jan 26, 2026 •

edited

Loading