Blockly AI Policy - Request for feedback

[Zoë Spriggs]

Hi everyone,

Over the past few months, the Blockly team has noticed an increased number of contributions that are partially or fully written by AI tools. In response to this, we have decided to write an brief policy regarding AI contributions to Blockly. 

Before we publish this policy, we’d love to get feedback from the community. You can review it below:

_____________________________________

As an open source library, Blockly is supported by external contributions in the form of Pull Requests (PRs), which are requests to merge code into the Blockly repository. 

The following document outlines our policy regarding the use of AI tools when making contributions to Blockly. 

Creating guidelines on the use of AI allows us to:

• Maintain a high-quality codebase

• Reduce the review burden on the Blockly team

• Provide accessible opportunities for new contributors

Human in the loop

PR descriptions and GitHub comments should be written by a human unless you need AI translation support. AI tools often produce verbose descriptions even for small changes, which makes them difficult to review. Writing the PR description yourself also ensures that you fully understand the code you’ve submitted. 

Humans are reading and reviewing all of the PRs that you submit, so we ask that you make it easy for a human to understand what your code does and why.

Good first issues

We label some GitHub issues as `good first issue` for developers who are new to Blockly. These issues are intentionally beginner friendly, well documented, and self-contained. 

The purpose of solving a `good first issue` is to learn about Blockly and the workflows that we use. Therefore, AI should not be used to write code or PR descriptions when solving a `good first issue`. If you have questions about an issue, you can always leave a comment on the issue, and the Blockly team will be happy to assist you. 

Enforcement

Pull requests should be more beneficial to the library than the time they take to review. The Blockly team may close PRs that violate this policy or are otherwise too lengthy to review efficiently. 

_____________________________________

Let us know your thoughts! 

Best, 

Zoë Spriggs

Blockly DevRel Engineer

[Alan Smith]

While I think these are good, I would add 3 things.

1) The human submitting the PR is responsible, so you must not list AI as the author or co-author.

2) You need to personally review all code before submitting a PR.

3) answering questions or comments brought up in the review of the PR must be made by the human submitter and not an AI agent

Just my 2 cents....

--Alan

--
You received this message because you are subscribed to the Google Groups "Blockly" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blockly+u...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/blockly/bc533502-30be-447b-8662-d2e8a8ad4addn%40googlegroups.com.

[Mark Friedman]

On Thu, Sep 10, 2026 at 9:08 AM Alan Smith <al...@randomsmiths.com> wrote:

While I think these are good, I would add 3 things.

1) The human submitting the PR is responsible, so you must not list AI as the author or co-author.

I would actually want to see the info that the AI authored or co-authored as prominently as possible and in the commit history, so I think it's good to have them listed.  That's independent of the issue of responsibility.  A human must be an author or co-author and should, obviously, be the responsible party.

-Mark

[Alan Smith]

Mark,

  I copied this from Kubernetes ( https://www.kubernetes.dev/docs/guide/pull-requests/#ai-guidance ) where they say that you must say it was developed with AI, but an AI agent can not be listed as author or co-author because a human has to be responsible.  

-Alan

--

You received this message because you are subscribed to the Google Groups "Blockly" group.
To unsubscribe from this group and stop receiving emails from it, send an email to blockly+u...@googlegroups.com.

To view this discussion visit https://groups.google.com/d/msgid/blockly/CAOk7GcToBwt%3DRtYmqbKmM0UAC%2BBouXJ_z79_W8nNJEAH1k3VVA%40mail.gmail.com.

[Mark Friedman]

I just don't see how an AI could be considered responsible or what even that means.  As long as a human is listed as an author along with the AI,  I don't see what the problem is. The human is the responsible party and that could be made clear in the policy.  Moreover, I think that the recording of the AI as an author/co-author on the commit is valuable information.

-Mark

-Mark

To view this discussion visit https://groups.google.com/d/msgid/blockly/CAEviOug9H31sit68iYLGn6HxOe2FDBcqPE9EhtWi6mpNq4D2GQ%40mail.gmail.com.

[Zoë Spriggs]

Hi Alan and Mark, 

Thanks for sharing your thoughts! From my understanding there are also legal reasons why some libraries are concerned about whether an AI should be listed as "co-author" or not, but of course the legal situation with AI contributions is rapidly changing along with everything else about AI :)

Regardless, it does seem that there's agreement on requiring that contributors explicitly disclose their AI usage. We'll add a section of the AI policy asking that AI usage be disclosed/explained in a couple sentences of the PR description. If we find that still isn't sufficient, we may add more specific requirements about exactly how to attribute AI in the future.

Best, 

Zoë