15:54:37 RRSAgent has joined #webmachinelearning 15:54:42 logging to https://www.w3.org/2026/11/12-webmachinelearning-irc 15:54:43 RRSAgent, make logs Public 15:54:44 please title this meeting ("meeting: ..."), anssik 15:54:45 Meeting: WebML CG Teleconference – 12 November 2026 15:54:49 Chair: Anssi 15:54:53 Agenda: https://github.com/webmachinelearning/meetings/blob/main/telcons/2026-11-12-cg-agenda.md 15:54:57 Scribe: Anssi 15:55:02 ScribeNick: anssik 15:55:10 Present+ Anssi_Kostiainen 15:55:15 RRSAgent, draft minutes 15:55:16 I have made the request to generate https://www.w3.org/2026/11/12-webmachinelearning-minutes.html anssik 15:58:43 sjdallst2 has joined #webmachinelearning 15:59:53 yoav has joined #webmachinelearning 16:00:25 Present+ Samuel_Dallstream 16:00:34 Present+ Umar_Iqbal 16:00:34 Mark_Foltz has joined #webmachinelearning 16:01:06 Present+ Benjamin_VanderSloot 16:01:16 ABN has joined #webmachinelearning 16:01:48 kush has joined #webmachinelearning 16:02:11 Anssi: as a reminder, we'll use IRC-based queue management in this meeting: 16:02:17 -> https://www.w3.org/guide/meetings/zakim.html#speakerqueue 16:02:21 Anssi: to suggest agenda topics, use Agenda+ label, e.g.: 16:02:26 -> Agenda+ https://github.com/webmachinelearning/webmcp/labels/Agenda+ 16:02:26 present+ 16:02:31 domfarolino has joined #webmachinelearning 16:02:33 present+ 16:02:33 BenGreenstein has joined #webmachinelearning 16:02:39 Anssi: welcome back everyone, we made significant progress at TPAC and have important follow-ups to discuss today 16:02:41 Julien has joined #webmachinelearning 16:02:44 Anssi: first, please welcome our latest new participants: 16:02:51 ... Christopher Wood and Thibault Meunier from Cloudflare 16:02:57 ... Artur Serra from Snyk 16:03:00 ... Taylor Dolezal from Dosu 16:03:05 ... Gary Klassen from Google 16:03:09 ... Liad Yosef from Ora 16:03:14 ... Peter Rosemann from Alpin Insight Solutions 16:03:20 ... Prabhu Rangaraju from Penske Logistics 16:03:24 ... Marinoni Moretto, Song Zhang, Daniel Mantei, Steven Burke, Chris McKenzie, Ragdan Abdulbaqi, Sai Vineela Ganti, Ishan Anand, Damisile Ayoola, Dongxun Jin, Juliano Sig, Caio Gondim, Damisile Ayoola, Dongxun Jin, Caio Gondim as individual contributors 16:03:31 ... welcome all on board this Community Group 16:03:42 Isaac has joined #webmachinelearning 16:03:45 Topic: Beyond WebMCP 16:03:48 fbeaufort has joined #webmachinelearning 16:03:51 Idan has joined #webmachinelearning 16:03:53 Anssi: this is a follow up to the TPAC Agentic Web breakout session where we discussed the steps beyond WebMCP: 16:04:00 -> Breakout minutes https://gist.github.com/yoavweiss/211c446c914a5abf03548652e31de9a8 16:04:02 noamr has joined #webmachinelearning 16:04:05 -> Dom's slides https://drive.google.com/file/d/1UO_UwtxtxZHqWH1B3Ou7_6Ftmd-X_uPb/view 16:04:16 -> Tentative Deliverables https://w3c.github.io/charter-drafts/2026/agenticweb-wg.html#incubation 16:04:46 Anssi: the goal of today's discussion is to advance the following proposed Agentic Web WG's tentative deliverables toward an explainer form: 16:04:52 ... - Agent Identify 16:04:58 ... - Agent Delegation 16:05:02 ... - Agent Interaction Measurement 16:05:13 ... the "Agentic Interaction Modes" deliverable needs some clarity wrt normative vs non-normative aspects 16:05:23 plh has joined #webmachinelearning 16:05:25 Anssi: for context, about Working Group vs Community Group interplay 16:05:41 ... when W3C charters a Working Group, the established pattern is to use a Community Group to handle incubation for topics related to the Working Group 16:05:53 ... this WebML CG serves that incubation function for WebMCP and related topics 16:06:18 ... W3C WG charters are expected to be well-defined wrt normative deliverable scope to allow W3C members evaluate IP commitments 16:06:42 ... to that end, this Community Group will produce explainers for the 3 proposed tentative deliverables (Agent Identify, Agent Delegation, Agent Interaction Measurement) to allow the Agentic Web WG charter to make timely progress to the Advisory Committee review 16:07:02 ... the explainers and draft specs developed in this Community Group are expected to advance to an appropriate WG after successful incubation 16:07:30 ... in a nutshell, the W3C Working Groups are governed by the W3C Process that applies the RF IPR licensing to the resulting specs, the end product aka W3C Recommendation 16:07:44 ... future proposals will advance through a Community Group incubation process similarly to WebMCP before moving to a WG for standardization and eventual W3C Recommendation 16:08:00 ... to make this work, seamless collaboration between this CG and the proposed Agentic Web WG is required to ensure a smooth transition 16:08:33 Anssi: the explainers should follow the usual structure, concise overview of the problem, proposed solution(s), use cases, alternatives, any known considerations wrt a11y, i18n, privacy, security 16:08:37 -> "Explainer explainer" document https://w3ctag.github.io/explainer-explainer/ 16:09:24 Anssi: with that backdrop, let's discuss the proposed tentative deliverables, identify owners who take responsibility for transforming the issues into explainers to allow the Agentic Web WG charter advance in a timely manner 16:09:36 ... I'll create repos for each proposed tentative deliverable to facilitate focused development and collaboration 16:09:40 ... any questions about this general goal? 16:09:54 q+ 16:10:03 ack Mark_Foltz 16:10:22 Mark: is the goal to have explainers for these topics to have them completed by the end of the month? 16:10:25 Anssi: yes 16:11:04 q? 16:11:19 Subtopic: Agent Identity 16:11:30 Anssi: issues #54 and #365 16:11:30 Issue 54 not found 16:11:30 Issue 365 not found 16:11:52 ... Agent Identity covers a mechanism for sites and agents to convey and negotiate identity with specific attribution of agent-originated interaction, for example: 16:12:02 ... "I'm an agent working on a user's behalf" 16:12:17 ... "This Submit event was initiated by an agent acting on behalf of a user" 16:12:30 q+ 16:12:37 domfarolino has joined #webmachinelearning 16:12:40 ... one proposed solution is to extend the Event interface with a new agentInvoked attribute, good starting point for the explainer 16:12:44 q? 16:12:45 -> https://dom.spec.whatwg.org/#interface-event 16:12:46 q+ 16:13:01 Anssi: Ilya explained the use cases and requirements for the proposed mechanism from Shopify's perspective drawing from large scale deployments, this additional good input for the explainer document 16:13:05 -> https://github.com/w3c/strategy/issues/576#issuecomment-6413882762 16:13:06 https://github.com/w3c/strategy/issues/576 -> Issue 576 [wg/agenticweb] Agentic Web Group Charter (by plehegar) [AI] [Horizontal review requested] [Security review completed] [In Charter Refinement] [charter] 16:13:12 q? 16:13:15 ack yoav 16:13:26 ryan-roemer has joined #webmachinelearning 16:14:06 Yoav: essentially I think it would be good to start with an explainer that outlines the problem, them try to map the solutions to the use cases and see what other groups are doing in e.g. IETF 16:14:23 q? 16:14:40 ack domfarolino 16:14:51 +1, we need to focus explainers on problem definitions and not solutions in the short time frame 16:15:16 Dom: I agree, we need to understand the problem we want to solve, and when the agent would expose its identity, e.g. through clicks 16:15:17 bvandersloot has joined #webmachinelearning 16:15:23 q? 16:15:26 q+ 16:16:07 ... I think we want a clearer picture what interface the agent will use to interact with the browser 16:16:56 ... we want to figure out the problem we want to solve and interface by which the agent interfaces with the browser 16:16:59 q? 16:17:01 ack bvandersloot 16:17:46 Benjamin: I wanted to take a step back and ask, to bring them in as WG items we need to recharter, I'm questioning why to rush this work, perhaps more natural incubation without rush is wiser 16:17:47 q? 16:17:55 q+ 16:17:57 q+ 16:17:59 ack Mark_Foltz 16:18:10 q+ Ehsan 16:18:26 Mark: I raised this as an issue against the draft, I wanted to ask what exactly is an agent and what technologies are in scope 16:19:23 q? 16:19:30 ack yoav 16:19:36 Ehsan has joined #webmachinelearning 16:19:48 q+ 16:20:18 Yoav: I agree with Ben's comment 16:20:58 ... I do think it is better to tackle these sooner than later, we don't want to rush to solutions, we need a clearer sense of problem statements 16:21:04 q? 16:21:13 ack Ehsan 16:22:07 Ehsan: I agree with Ben, Yoav and Mark, this discussion needs a lot of pre-discussion before getting to solution, I'm fine to have a document that defines the problem, also parallel discussion expected in WebAuth and elsewhere 16:22:24 q? 16:22:46 happy to help with this 16:23:08 gb, this is webmachinelearning/webmcp 16:23:08 anssik, OK. 16:24:37 I can help with mapping use cases for the explainer 16:24:57 +1 16:25:05 +1 16:25:08 RESOLUTION: Yoav and Dom to draft an explainer for Agent Identity, focusing on problem statement and use cases. (issues #54 and #365) 16:25:13 Subtopic: Agent Delegation 16:25:17 Anssi: issues #44 and #54 16:25:29 ... mechanisms for users to grant agents authority to act on their behalf on a site to take actions 16:25:41 ... relevant in this context is to understand implications for different interaction modes 16:25:58 ... including enterprise use cases and requirements 16:26:11 ... two facets of agent delegation: action-specific permissions and agents as separately authorized clients 16:26:19 ... 1) action-specific permissions, issue #44 16:26:32 gb has joined #webmachinelearning 16:26:32 ... "I trust agent 'Foo' to take this action 'Bar' in the future on this site (example.com), don't ask me again" 16:26:41 q+ Khushal 16:26:52 ack Khushal 16:27:49 Khushal: both agent and the site want to ensure they have explicit user's consent, avoid duplicated consenting once for browser, once for agent 16:29:09 Umar: a complicated topic, for AI agents, one of the ideas is to propagate contextual information 16:29:12 Vineelaganti has joined #webmachinelearning 16:29:32 ... a lot of work on checking user's prior interaction, are they confident making a new decision based on their history 16:29:46 ... or what not giving a permission implies 16:30:06 ... 2) agents as separately authorized clients, issue #54 16:30:07 https://github.com/webmachinelearning/webmcp/issues/54 -> Issue 54 Challenging assumptions of Identity within WebMCP (by EmLauber) [beyond-scope: agents-on-the-web] 16:30:22 ... "agent 'Foo' is authenticated and authorized to call WebMCP tools on this site (example.com)" 16:30:28 q+ 16:30:36 ... Emily opened this issue a year ago, perhaps Khushal or Victor can speak to this? 16:30:40 ack Mark_Foltz 16:31:15 frmoretto has joined #webmachinelearning 16:31:38 Mark: I agree that identity and permission are linked, makes sense to work on them together in an explainer form 16:31:39 q? 16:31:41 q+ 16:31:50 Mark: permissions will build on top of agent identity work 16:31:51 q? 16:31:54 ack BenGreenstein 16:32:22 Ben: agree, part of agent identity is proof it is acting on behalf of specific user 16:32:22 q? 16:32:43 Anssi: I see Victor made a connection to IETF's work on Web Bot Auth (webbotauth), is this relevant for our discussion? 16:32:47 -> https://datatracker.ietf.org/wg/webbotauth/about/ 16:33:07 present + Sai_Vineela_Ganti 16:33:14 Yoav: I talked to one of the IETF WG chairs, it seems aimed at server-side communication 16:33:33 ... on the client-side, no way to keep private keys secure, seems less relevant here 16:33:35 q? 16:33:38 q+ 16:33:52 Anssi: and in context of charter review, Mark made a comment about the agent delegation scope: 16:33:55 -> https://github.com/w3c/charter-drafts/issues/906 16:33:56 https://github.com/w3c/charter-drafts/issues/906 -> Issue 906 [wg/agenticweb] Agent Delegation deliverable scope could be broadened (by markafoltz) 16:34:25 Anssi: Mark's comment: "The scope of agent delegation should extend across agent interaction modes, including installable agents, and agents running as ordinary web content or cross-origin frames. For agents to be truly useful, they also need other capabilities, such as the ability to observe page content, discover and invoke tools, and simulate UI actions." 16:34:36 Anssi: is this still open? 16:35:00 Mark: I haven't looked at the latest, if this has been addressed, I wanted to soften the language to see what capabilities make sense 16:35:26 ... I can follow up on this detail 16:35:28 q? 16:36:01 Mark: I'm referring to the scenarios presented in Dom's slides 16:36:06 q? 16:36:10 ack bvandersloot 16:36:31 https://datatracker.ietf.org/group/audit/about/ 16:36:38 q+ 16:36:52 Benjamin: IETF proposal relevant here is Agent Use of Delegation and Interaction Traceability (audit) 16:36:57 ack yoav 16:37:29 Moderation Of unLinkable Endorsement 16:37:42 Yoav: another forming group BOF'd enables bots or agents to pass token acquired by user to prove that they're attached to an actual user rather than bots operating independently 16:37:50 q? 16:37:50 I think MoLE was for anti-fraud purpose though 16:38:00 Porque no los dos? 16:38:05 q? 16:38:10 https://datatracker.ietf.org/doc/bofreq-thomson-moderation-of-unlinkable-endorsements-mole/ 16:38:34 some of the identity use cases are very much related to anti-fraud 16:39:52 Johann: my team might be interested in working on Agent Delegation 16:40:02 Johann has joined #webmachinelearning 16:40:23 I think the context for MoLE is that many of the new assistants (Muse etc.) are using a cloud VM for all their users, acting as a proxy. traditionally they got blocked and they now need to prove they are acting on behalf of many users 16:42:11 +1 16:42:14 +1 16:42:21 +1 16:42:22 RESOLUTION: Johann and Mark to draft an explainer for Agent Delegation. (issues #44 and #54) 16:42:29 Subtopic: Agent Interaction Measurement 16:42:33 Anssi: issue #186 16:42:34 https://github.com/webmachinelearning/webmcp/issues/186 -> Issue 186 WebMCP "real-user measurement" (by yoavweiss) [beyond-scope: agents-on-the-web] 16:42:37 Anssi: APIs for pages to measure in-field agent interaction performance, reliability 16:42:44 ... Yoav shared a strawman proposal building upon the PerformanceObserver spec with new entries for agent interaction measurements 16:42:49 -> https://github.com/webmachinelearning/webmcp/issues/186#issuecomment-6369314489 16:42:50 present+ Sai_Vineela_Ganti 16:42:50 https://github.com/webmachinelearning/webmcp/issues/186 -> Issue 186 WebMCP "real-user measurement" (by yoavweiss) [beyond-scope: agents-on-the-web] 16:42:56 Anssi: Yoav, could talk to your strawman proposal? 16:43:05 Anssi: I think this strawman proposal is a good starting point and would feed into to the explainer 16:43:15 ... Yoav, could you take this forward and draft the explainer for Agent Interaction Measurement? 16:43:23 q+ yoav 16:43:26 ack yoav 16:44:25 Yoav: PerformanceObserver that monitors tool calls so that would avoid the need to wrap tool calls to count and measure them, also gives us some form of interaction ID to enable us get more visibility into agentic turns 16:44:45 ... it might also be interesting to connect this to user interaction 16:44:46 q? 16:45:02 ... connect tool calls with UI updates and think how that is to be represented 16:45:14 q+ 16:45:40 ... tool call is WebMCP specific, agent interaction entries could give us an overview whether tools are used or not, how long user interaction takes from the user's perspective 16:46:35 ... it would be interesting to know the outcome of those interactions, user's giving thumbs up or down interaction to agent, agent could assess whether the missions was accomplished, abandoned 16:46:36 q? 16:46:58 Yoav: many entries are optional 16:46:58 q? 16:47:01 ack domfarolino 16:47:40 Dom: it sounds like this task is extracting information from agent, probably come up with a giant draft and run it through agent vendors to see what metrics would be helpful to them 16:47:44 liad has joined #webmachinelearning 16:47:51 q? 16:47:52 q+ 16:48:05 Yoav: sounds good to me 16:48:28 Dom: the first stab is more like here are the useful metrics 16:48:32 ack liad 16:48:55 Liad: good idea, in the MCP world we are standardizing intent and feedback, maybe cross-pollinate on these ideas 16:50:22 Yoav: we can outline the use cases and rough skeleton, bits of information we need from agents and review with agent providers 16:50:53 Sai Vineela(Target, enterprise UI/a11y): +1 to measuring outcomes from the user's perspective. Suggest also covering human-oversight signals: did the user intervene, correct or undo an agent action, was a confirmation shown/accepted, and was the agent-caused UI change perceivable (visible update, a11y announcement). 16:51:44 +1 16:51:59 +1 16:51:59 +1 16:52:01 RESOLUTION: Yoav to draft an explainer for Agent Interaction Measurement, outline the use cases and rough API sketch, bits of information we need from agents and review with agent providers and privacy group. (issue #186) 16:52:08 Subtopic: Agentic Interaction Modes 16:52:17 Anssi: this deliverable was initially envisioned as a non-normative guidelines doc for agentic interaction in different modes, browser, extension, automation, headless 16:52:27 Anssi: Dom's review comments notes this has normative aspects: 16:52:27 ... "a gateway from some defined "agent" into the web platform to normatively expose this from a truthful agent harness." 16:52:30 -> https://github.com/w3c/charter-drafts/pull/901#discussion_r4411224898 16:52:31 https://github.com/w3c/charter-drafts/pull/901 -> MERGED PR 901 [wg/agenticweb] Generalize WebMCP deliverable; add identity, delegation, measurement and interaction modes (by yoavweiss) 16:52:34 q+ Dom 16:52:39 frmoretto has joined #webmachinelearning 16:52:42 Anssi: we need to clarify which parts are normative and which are non-normative, and for the normative aspects, have a reference 16:54:03 Yoav: going back to Ben's comments, no rush to go into this particular document, others are higher priority 16:54:11 +1. It also feels like an extension of "Agent Identity" (not sure if that was already mentinoed) 16:54:36 q? 16:54:39 ack Dom 16:56:04 _1 16:56:06 +1 16:56:19 +1 16:56:21 RESOLUTION: Merge Agentic Interaction Modes considerations with Agent Identity. 16:56:25 Topic: Agentic Web Threat Model 16:56:29 Anssi: I'm pleased to welcome our guest speaker Umar Iqbal, Assistant Professor, WashU 16:56:36 ... Umar's research will bring new perspectives into the Agentic Web Threat Model deliverable 16:56:47 ... Umar will present novel research on a framework for more secure agentic web interactions, please take it away, Umar 16:56:51 -> Paper https://arxiv.org/pdf/2610.22697 16:56:54 -> Code https://github.com/llm-platform-security/intact 16:56:58 -> Data https://huggingface.co/datasets/umariqbal/web-agent-attack-benchmark 16:58:07 [ Umar presents slides: Securing Agentic Browsing Interactions with Web Semantics ] 16:58:57 RRSAgent, draft minutes 16:58:58 I have made the request to generate https://www.w3.org/2026/11/12-webmachinelearning-minutes.html anssik 17:02:01 Anssi: let's find a time to get Umar back to present on this research again 17:02:19 Topic: WebMCP 17:02:24 s/Topic: WebMCP// 17:02:29 RRSAgent, draft minutes 17:02:30 I have made the request to generate https://www.w3.org/2026/11/12-webmachinelearning-minutes.html anssik 17:04:08 Present+ Ali_Spivak 17:04:21 Present+ Andrei_Besleaga 17:04:33 Present+ Ben_Greenstein 17:05:13 Present+ Chris_Harrelson 17:05:35 Present+ Dominic_Farolino 17:05:45 Present+ Ehsan_Toreini 17:06:01 Present+ Francois_Beaufort 17:06:22 Present+ Johann_Hofmann 17:06:29 Present+ Idan_Levin 17:06:38 Present+ Isaac_Ahouma 17:06:48 Present+ Jan_Williams 17:07:03 Present+ Jeff_Whelpley 17:07:18 Present+ Julia_Pagnucco 17:07:41 Present+ Julien_Bataille 17:07:57 Present+ Khushal_Sager 17:08:14 Present+ Mark_Foltz 17:08:34 Present+ Philippe_Le_Hegaret 17:08:46 Present+ Tom_Jones 17:09:04 Present+ Yoav_Weiss 17:09:10 RRSAgent, draft minutes 17:09:11 I have made the request to generate https://www.w3.org/2026/11/12-webmachinelearning-minutes.html anssik 17:12:36 s/issues #54 and #365/issues https://github.com/webmachinelearning/webmcp/issues/54 and https://github.com/webmachinelearning/webmcp/issues/365 17:12:38 RRSAgent, draft minutes 17:12:39 I have made the request to generate https://www.w3.org/2026/11/12-webmachinelearning-minutes.html anssik 17:13:03 s|issues #54 and #365|issues https://github.com/webmachinelearning/webmcp/issues/54 and https://github.com/webmachinelearning/webmcp/issues/365 17:13:04 RRSAgent, draft minutes 17:13:06 I have made the request to generate https://www.w3.org/2026/11/12-webmachinelearning-minutes.html anssik 17:14:01 s/this additional/this is additional 17:14:15 s/them try/then try 17:15:40 s/the draft/the draft charter 17:17:58 s/... "agent 'Foo' is/Anssi: "agent 'Foo' is 17:19:51 s/could talk/could you talk 17:21:01 s/the missions was accomplished, abandoned/the mission was accomplished or abandoned 17:25:11 s/find a time/find additional time 17:25:19 RRSAgent, draft minutes 17:25:21 I have made the request to generate https://www.w3.org/2026/11/12-webmachinelearning-minutes.html anssik 17:28:26 Present+ Ryan_Roemer 17:29:16 Present+ Francesco_Marinoni_Moretto 17:29:51 RRSAgent, draft minutes 17:29:53 I have made the request to generate https://www.w3.org/2026/11/12-webmachinelearning-minutes.html anssik