W3C

– DRAFT –
RQTF meeting

27 August 2025

Attendees

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

Meeting minutes

Accessibility of Machine Learning and Generative AI.

jasonjgw: work is ongoing, anything to minute or discuss?

janina still behind in incorporating Scott's work

janina: we should probably make sure agentic AI is incorporated somewhere.

<janina> Model Context Protocol: https://en.wikipedia.org/wiki/Model_Context_Protocol

JPaton: As I understand it Agentic AI are AI Agents that can be given a task and they will interface with other software and APIs and will come back with a response or having performed a function

jasonjgw: a large part of a presentation I saw recently from Greg Vanderheidan was about Agentic AIs Acting on behalf of a user to carry out tasks. If APIs and systems were designed to be usable by agentic AI then it could work on behalf of people with a wide variety of disabilities.

jasonjgw: the reports I'm seeing in the media suggest that agentic AI will be the predominant form of AI in the future.

jasonjgw: given the structure for this publication, what do we want to say about agentic AI and where?

janina: W3C is heavily associated with APIs, there may be a meta-API conversation that needs to be had here.

https://www.theregister.com/2025/07/12/ai_agent_protocols_mcp_a2a/

JPaton: I've added a link to an article talking about MCP and A2A which is a google API for agent to agent communication. In my head the Agentic APIs would be linked to the Web of Things work.

jasonjgw We could, in the document, highlight high level ways that agentic AIs could support accessibility work without getting into which APIs would be used.

janina: maybe this could go to WCAG 4 and deal with the topic in a taxonomical way.

jasonjgw: If agentic AI can work with websites, for example, on behalf of a user then some of the assumptions around designing for accessibility could go away.

janina: That kind of screen scraping may be going away and reports are that currently it's not a robust mechanism for getting reliably information from the page.

scott_h: there are questions of timing and whether this affects our intended schedule.

jasonjgw: but it would be useful to get those placeholders in there.

Miscellaneous topics.

jasonjgw I gave my presentation yesterday on AI and ML at Monash University. There were some good questions afterwards and some interested prototypes presented by the students around tactile displays, natural language and AI.

jasonjgw: they have a series of projects on Sign Language too.

https://jasonjgw.net/presentations/technology-disability-intersection-slides.html

Minutes manually created (not a transcript), formatted by scribe.perl version 244 (Thu Feb 27 01:23:09 2025 UTC).

Diagnostics

Maybe present: JPaton

All speakers: janina, jasonjgw, JPaton, scott_h

Active on IRC: janina, jasonjgw, JPaton, scott_h, stacey