W3C

– DRAFT –
RQTF meeting

19 August 2026

Attendees

Present
janina, jasonjgw, John_Paton, stacey
Regrets
-
Chair
jasonjgw
Scribe
jasonjgw, janina, stacey

Meeting minutes

Accessibility of Machine Learning and Generative AI draft.

Jason: anythign we can usefully discuss and coordinate on this?

Janina: No, I've started to build new file based on outline TOC that was shared and commented on. As discussed last week, moving to W3C style HTML markup in git. Need to trim down what was in there. But the AI helped clarify what we were thinking. GOing to try to capture that. And then copy over what we already have in existing doc and decide if

ready to merge and make the new doc the main doc. Won't wait for all three to be done to bring it back, but need to get first level done to review.

Jason: when would you like it to come back to a meeting?

Janina: let's try for next week

Jason: Doesn't mean you need to have anything...

Jason: anything else on this item?

Game accessibility.

Jason: John circulated an update on the mailing list. Thank you for that. Updated declares a draft of gap analysis complete and text of a cover note to APA working group

John: That's a good summary. Did anyone have a chance to look at it? Evaluation and proposal section at the end were main changes.

Jason: On basis of that, already suggested analysis was in right direction. Haven't given it another review yet. Provisionally saying developing in the right direction. Any reactions to John's status report?

Janina: Please you've gotten it to this point and ahead of the calendar. Seriously pleased. I haven't read it, but perked up and now is the time to take a serious look at it. Will respond on list if anything significant. But have a question and come back to it in a minute. Want to talk timeline. Not a good time to do a CFC right before labor day

weekend (sept 7). But next Wed is out last Wed in August. If the doc is well put together, then that will exactly fit that timeline, a call for consensus with APA right after labor day. Should have a decision - usually about a week. Is there in cover documentation or cover letter a statement - short work statement...here's the gaps and here's what

we propose? Anything in there like that already?

John: one sentence, but recommending the intentional friction aspect...I don't know if it's as succinct, there's a purpose at the beginning.

Janina: i would like it to flag, useful ways to look at what type of breakage we would accept and under what circumstatnces and building the GAUR and develop concepts futher.

John: Happy to encapsulate that and make changes

Jason: Second what Janina is saying. Draft abstract would be useful

John: proposal at the end, methodology and intentional and unintentional friction...Breaking it down like that.

Janina: Test might be accessible guidelines if they want to pick it up, but developing that this exists and why important and how we used it in the GAUR and where we're landing...Incorporate in WCAG 3...game plan at top of my head.

Jason: Sounds reasonable to me. Having characterized appropriate ways of analyzing the problems is as much as we can achieve...elaborate in a way useful to people who are developing guidelines and practionioners in field encountering these...

Janina: if can get into WCAG 3, then a path to regulators to pick up on

Jason: John - bring this back next week?

John: yes

Jason: slight elaboration on scope of work and then were' fine. Comments may come in on analysis and where it stands

Miscellaneous topics.

Jason: other things going on we should know about?

Janina: nothing specific from me. Didn't have the XR...no planning meeting for APA.

Jason: Let's introduce taht when it's appropriate. bring back here when we have clarity

Janina: i'll give a heads up coming early sept. and have a proposal.

jason: on XR issue, raise that at appropriate time when we get clarification..discuss if needs discussing

janina: yes

Jason: Articles and presentations from open AI regarding recent security incident that had media publicity. Has to do with cooperation and on cooperation between agents. They can cooperate to breach securtiy to perform tasks or cheat (as commentators put it) and communicate among themseelves using mechanims you didn't expect. Move into a world

where complex interactions between different users LLM agents and when that invovles humans as well can complicate the issue of oversight and potentially open opportunities for a11y problesm to be solved but cooprartion among multiple agents and haven't thought about those implicataions

janina: probably add that into the paper. Who decides..."my agent" - if my agent I should ahve some say in that

jason: interacting with other agents, what sort of control do you have with that, and benefits if solvin your or my a11y issue along the way

Janina: who indemnifying....fiduciary implications. unavoidable

jason: same person could invoke multiple agents working on same task or parts of task and multiple agents work on it

Janina: and compete for my approval

Jason: playing out in research right now and head our way in the next couple of years
… slight gloss on agent issue with multipl cooporating agents...

Janina: authorized and unauthorized

Jason: yes, that's what people getting concerned about

Janina: browser decided is not sufficient in my view

Jason: agent to perform a long process could access web applications and performing actions you don't know about
… indirectly have a11y implications. Explain users role may need to be to oversee these types of systems and user's agent may seek to make an interaction more accessible could interact with other agents. so my agent could interact with the airline's agent in the future

Janina: think about what we might need to say about that

Jason: probably not a lot, but acknowledge that it will happen

Janina: would hate if yet another set of archane and complicated terms of service...those things that everyone doesn't read

Jason: possible as well

Janina: i think legal situation with meta will have profound impact on industry and won't be as freewheeling as some of those judgements come down.
… social web might be called the anti-social

Jason: if my agent and works with airline agent and books me on wrong flight am i responsible?

janina: tried to book with innosearch it booked me with the wrong class
… it did not do what it said. and innosearch (bug report) i got sent to their help pages

Jason: issue for us, if anyone has a cognitive disability that makes dealing with these issues more difficult...will be at a disadvantage

Janina: disability tax at all levels

Jason: error handling that could be more difficult

Jason: multiple agent world...what happens when it doesn't work as intended. Might have to discuss this

janina: when it breaks. Because it will

Minutes manually created (not a transcript), formatted by scribe.perl version 248 (Mon Oct 27 20:04:16 2025 UTC).

Diagnostics

Maybe present: Jason, John

All speakers: Janina, Jason, John

Active on IRC: janina, jasonjgw, JPaton, stacey