– DRAFT –
ATAG CG Meeting

date: 2026-02-27

Charter Draft

w3c-cg/atag#28

<Andrew> Wendy: For today, lets talk abou the first draft of the charter.

<Andrew> Wendy: First attempt to put down the rought idea of what we want to do

<Andrew> Wendy: Charter less chaotic when you take out helper text

<Andrew> Wendy: Updating references to 2.2 and any other references to older things

<Andrew> Wendy: updating with LLMs changing to needs of users

<Andrew> Andrew: Wendy: Out of scope: what we wont' be doing. Not doing content standards. We will reference WCAG. Our deliverables: Incubate ideas for the next version which will ead us to a working group charter

<Andrew> Wendy: Another document I call the gap analysis, what is the difference between 2.0 and what we need and can we address them

<Andrew> wendyreid: We will cooperate with other groups.

<Zakim> Charles, you wanted to clarify non-normative report format. is this specifically a note?

<Andrew> Charles: The non-normative reports: are those report deliverable something that can live in google doc or wiki; should it specify the format

<Andrew> wendyreid: we have full control.

<Andrew> wendyreid: they call them community group reports but they are essentially notes

<Andrew> shivaji: once its published its available correct? wendyreid: yes its published and available.

<Andrew> charles: do the reports require horizontal review?

<Andrew> Andrew: LLMs provide great note taking :)

<Andrew> wendyreid: the research report is something I gleaned form ASC

<Andrew> jutta: its to provide reference to what has not been done, or create clauses to other standards

<Charles> some of the early Silver Task Force work was research and synthesizing it to insights. some of those insights were published as google docs linked to from the wiki – but never published formally.

<Andrew> jutta: it opens the lines of negotiation.

<Andrew> jutta: its the basis of harmonization and contextualization of what we are proposing

<Andrew> sambhavi: can help with iso integration (sorry for this summarization; adhd moment)

<Andrew> wendyreid: anthing I missed?

<Andrew> simonrjones: saw a comment from heather in reference to ai being used. headless cms being used.

<Andrew> nicole: my one and only committee Ive worked on. What are the other examples and samples I could use that would be useful

<Andrew> jutta: I noticed in github about a greater specificity of AI. Dont' only limit to generative AI. There are tools that are AI, but not generative in testing and coding and a variety of other things

<Charles> always fascinated by the pros and cons of specificity and ambiguity in labels and taxonomies

<Andrew> wendyreid: I used this as an example in the hope we will move on to a working group charter, same language, same rigour.

<Andrew> wendyreid: I thought it was helpful to point people to the work we do; a signal to other people to whate we're working on

<Andrew> nedzimmerman: just to follow on jutta point on specificity. I think its a really good point. As far back as 2019 I saw things that we are doing that aren't LLM such as alt text generation

<Andrew> wendyreid: there are LLMs but automation

Andrew: We're doing the same, automation and generative AI

jutta: there are ai taxonomies and we can talk about ai authoring and how its being used; an informative document

<Charles> +1 to taxonomy of AI

https://github.com/w3c-cg/atag/wiki/Use-Cases-for-Authoring-Tools

https://aimac.ai

wendyreid: we are used to release notes. With AI thats not what happens. We've done these things. They give percentages. It does y percent better.

<Charles> AI release notes are pure marketing

wendyreid: we dug into gemini. Instead of a test suite do we create a benchmark?

https://agi.safe.ai

https://storage.googleapis.com/gweb-uniblog-publish-prod/original_images/gemini_3-1-pro__benchmarks.gif

wendyreid: I will go through comments

Evelyn: is there a definition of "authoring tool"

Any application(s) that can be used by authors (alone or collaboratively) to create or modify content for use by people (authors or end users).

Charles: do we still have a dependency on other chartering groups like AG.

wendyreid: we won't have that awkward dependency

jutta: to comment on benchmarks and definition on authoring tools. One of the things we aligned on with ASC was to align on user interfaces

nedzimmerman: I came across this, I'll put it in the chat. There seems to be an attempt to get what llm accessibility means

https://conesible.de/wab/whitepaper_webaccessbench.pdf

wendyreid: there is another sister project to AMAC

https://microsoft.github.io/a11y-llm-eval-report/

wendyreid: we will spend time on charter comments

Andrew: Curious, after the charter, what happens?

https://www.w3.org/TR/wcag3/

https://conesible.de/wab/