[ai-control] Re: -04 drafts for evaluation

[Jo Levy <[e-mail]>]

I echo Timid Robot’s and Alissa’s comments about examining whether there is a false dichotomy, and I second Timid Robot’s suggestion that we look at the issues already raised but not addressed, such as #155.

In my comments to Issue 155, for example, I pointed out that part of the problem ( as I see it) is that the top level category is not framed in the same way as the subcategories.  Both subcategories describe—and start with—“the act of using an asset”.  But the top level category starts with “the act of using automated processing on one or more assets.”

Whether the top level is called AI processing or automated processing, the subcategories talk about the purpose for which the asset is used while the top level talks about the type of action performed on the asset:

4.1:  “The act of using automated processing on one or more assets…”
4.2:  “The act of using an asset to train or fine-tune a foundation model.”
4.3: “The act of using an asset in an AI-based system to generate outputs that are presented to clients of that system.”

To correct this, 4.1 should also start with “the act of using an asset.” For example:

4.1: The act of using an asset in an AI-based system or to train or fine-tune an AI-based system.

That doesn’t solve the problem that many have raised that selecting All AI=n would likely have unintended consequences of preventing most search, crawl, and internal uses that now or in the future will be rely on an “AI-based system.”

But it might help parse the issues and concerns amongst this group’s participants.

From: Alissa Cooper <[e-mail]>
Sent: Monday, November 3, 2025 4:07 AM
To: Timid Robot Zehta <[e-mail]>
Cc: Leonard Rosenthol <lrosenth=[e-mail]>; Farzaneh Badiei <[e-mail]>; Paul Keller <[e-mail]>; [e-mail]
Subject: [ai-control] Re: -04 drafts for evaluation

I agree about the false dichotomy and I think it would be good to explore Timid’s suggestion to figure out if there is a path there that can resolve concerns on multiple sides of the question about automated processing.

A different way of breaking through the false dichotomy would be to maintain a single vocabulary document but have it contain different branches or modules that are scoped to different use cases. The crux of the contention here seems to be that the web crawling world already has its own vocabulary (mostly explicitly defined but in some corner cases still contentious or undefined). The broader world of asset use for AI does not, or to the extent that it does, it is much more nascent. The needs differ between the wider set of asset use cases and the web use case. Trying to address the needs of both use cases with exactly the same definitions can’t overcome the disparity between the web world with its entrenched vocabulary and semantics and the wider world of use cases that lacks that legacy.

Alissa


On Nov 3, 2025, at 1:36 AM, Timid Robot Zehta <[e-mail]<mailto:[e-mail]>> wrote:

I continue to believe that inclusion or removal of the top level category is a false dichotomy.
The concerns about how broad it is can both be satisfied by reducing its scope. If those concerns are satisfied, it won't need to be removed and we can continue to benefit from its utility.

Echoing my proposal from 2025-09-02<https://mailarchive.ietf.org/arch/msg/ai-control/YQA1sJUZncVXqKQS4Uw1zuePv1I/> (though better documented in #155<https://github.com/ietf-wg-aipref/drafts/issues/155>), I think we should replace "Automated Processing" with "AI Processing".
To date, there has been no substantial discussion of this proposal.

On Mon, Nov 3, 2025 at 12:28 AM Leonard Rosenthol <lrosenth=[e-mail]<mailto:[e-mail]>> wrote:
Hey Farz….

I have answered this question for multiple times - in fact, I think at every single F2F meeting (and probably in email as well).

>why do we need an "all" or a "tdm" or an "automated processing" category when the bot can be blocked anyway at robots.txt level should they want to allow or block it?

Because this vocabulary is not just about bots and robots.txt​.  As discussed in a related thread, there are many other attachment methods that are waiting on this work - including IETF’s own HTML attachment method - that have nothing to do with bots or robots.

Leonard

From: Farzaneh Badiei <[e-mail]<mailto:[e-mail]>>
Date: Sunday, November 2, 2025 at 12:50 PM
To: Paul Keller <[e-mail]<mailto:[e-mail]>>
Cc: [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>
Subject: [ai-control] Re: -04 drafts for evaluation

EXTERNAL: Use caution when clicking on links or opening attachments.


Paul,

If Foundation Model Production replaces AI training and Generative AI training why not just use those terms? And in my opinion (and I have been trying to educate myself on this AI terms for the past two years so bear with me) Model Production is the stage after training and production  means producing something. So in a way I think Foundation Model Production is dealing with after training but in our case it includes both production and training. Which makes it broader.

I have another question, which we have raised a number of times so I apologize if you answered already: why do we need an "all" or a "tdm" or an "automated processing" category when the bot can be blocked anyway at robots.txt level should they want to allow or block it? I think it's important to have a clear cut answer. Krishna has thankfully raised the issue here again and he explains it better than me: https://github.com/ietf-wg-aipref/drafts/issues/178


I am working on my other comments and will send it to the list.


On Sun, Nov 2, 2025 at 12:23 PM Paul Keller <[e-mail]<mailto:[e-mail]>> wrote:
Dear all,

I also have a question of understanding.

@Farz – in your last message you wrote:

> On 1 Nov 2025, at 21:13, Farzaneh Badiei <[e-mail]<mailto:[e-mail]>> wrote:
>
> I have to think really hard about  replacing training generative AI with the Foundational Model Production category. Now that we have that change, then automated processing definition has become even more problematic. Too broad, it includes everything.

This echoes something Jo wrote earlier in this thread:

> On 30 Oct 2025, at 23:20, Jo Levy <jlevy=[e-mail]<mailto:[e-mail]>> wrote:
>
> Moreover, the change to the term "Foundation Model" instead of AI Training, makes it even more apparent that a top level category is not needed.


Could either of you help me understand why you think this is the case? I’m trying to follow the reasoning, but I’m not sure I do. In my reading, the new “Foundation Model Production” category simply replaces the previous “AI training” and “Generative AI training” categories. I don’t see how that change affects the rationale for having (or not having) a top-level category. But maybe I am missing something here?

Best, Paul

--
ai-control mailing list -- [e-mail]<mailto:[e-mail]>
To unsubscribe send an email to [e-mail]<mailto:[e-mail]>
--
ai-control mailing list -- [e-mail]<mailto:[e-mail]>
To unsubscribe send an email to [e-mail]<mailto:[e-mail]>
--
ai-control mailing list -- [e-mail]<mailto:[e-mail]>
To unsubscribe send an email to [e-mail]<mailto:[e-mail]>