@otyugh@pouet.chapril.org unless @gargron@mastodon.social has changed his mind in the preceding few months, he is against AI use in general.

Whether that extends to the rest of the core development team I couldn't say.

@julian Feels like a few ones already passed

Like that one https://github.com/mastodon/mastodon/commit/a47ed3104718ce6f69bca2c028b267988a28c26b

@otyugh@pouet.chapril.org <img class="not-responsive emoji" src="https://activitypub.space/assets/plugins/nodebb-plugin-emoji/emoji/android/1f44d.png?v=9f67c8a21b0" title="" /> suppose that only extends as far as his own commits then. Does this pose a problem for you?

I'm pro local LLM, so I regularly have it assist me in my own work. Although I am not sure whether you can call 35 billion parameters "large" nowadays.

@otyugh Mastodon has a stated AI contributions policy here:

https://github.com/mastodon/.github/blob/main/AI_POLICY.md

ActivityPub does not have an AI coding policy, because ActivityPub is an open protocol. You can use any implementation tools and techniques you want to build ActivityPub software. Nobody can tell you no.

@julian
>Does this pose a problem for you?

Yes I ideally want to do without developpers (and their software) using big generative models.

I don't care much about local LLM (I don't even understand why you'd bring it up). I feel like local LLM are mostly used as an excuse to say "not all LLM are evil, pace yourself" and stop at that. To me it's offtopic. It's not the general use, I'm talking about a societal issue, not about their niche counterpart that most people won't ever get to use.

@otyugh That said, I had a poll about this recently, and about 50% of respondents said it wasn't OK to run LLM-assisted software on the Fediverse.

https://cosocial.ca/@evan/116217855307998592

@evan activity pub : Interesting. I see where it's coming from I guess.

Mastodon : ouch, I feared as much...

@evan@cosocial.ca I think that AI policy is quite well-written, and understandable by humans — something lacking in a lot of legal-adjacent documents in general.

@julian I think that the Open Source community is converging on some policy points around AI-assisted contributions, among the projects that allow AI-assisted contributions:

- Disclosure: usually with Assisted-By: in Git commit message
- Responsibility: human is accountable for code
- Licensing: still need to assert that code meets project's license requirements

There's a good review by Redmonk from earlier this year:

https://redmonk.com/kholterhoff/2026/02/26/generative-ai-policy-landscape-in-open-source/

@otyugh there's a W3C copyright policy:

https://www.w3.org/copyright/

And one for patents:

https://www.w3.org/policies/patent-policy/

They are both permissionless.