[ig/webai] Web&AI Interest Group

[ruoxiran]

Evaluation

cc: @dontcallmedom
The rapid growth of AI technologies both inside and outside W3C increasingly impact the Web. Currently, there is no single, official place within W3C to coordinate broad AI-related discussions. Existing groups are either narrowly focused (like WebML WG) or fragmented across overlapping Community Groups. To address this, we propose establishing a Web & AI Interest Group to serve as a formal entry point for AI discussions—tracking trends, connecting groups, supporting workshops(e.g., web and AI Agent workshop), and exploring how AI agents protocols may influence the Web.

• Draft Charter

Draft Charter

Expected end of charter refinement phase: End of August

• Will this work help to lead the web to its full potential?

Yes. As I mentioned in the first paragraph, AI becomes embedded in Web applications, browsers, and other web infrastructure, there is a growing need for coordination across standardization efforts.

• Is the work Rec-track ready?

No, this is an IG, Deliverables will be AI-related reports.

• Do we have the ecosystem of participants needed to make the work successful?

Yes. The group could draw from a diverse ecosystem across AI and Web stakeholders:

1. Participants could come from W3C Members, including browser vendors, cloud providers, telecoms, academic institutions, and AI companies, and users (e.g., Google, Microsoft, ByteDance, Huawei, China Mobile)

2. Interest: Early support could come from participants in the W3C team who are interested in AI stuff, WebML WG, Webagent CG, AI Agent Protocol CG, etc

3. Related existing W3C web tech: WebNN, WASM, WebGPU, RDF/JSON-LD, AI APIs in WebML CG, DID, VC, etc

The group is also planning to coordinate with external orgs such as the IETF and ITU-T.

digging deeper:

Will we be able to make it succeed?

• right participants interested. What does the ecosystem look like?
  W3C Members from browser vendors, cloud providers, telecoms, academic institutions, AI companies, and users.
• implementation likely
  No implementation expected.
• manageably sized problem
  Need to find Chairs to support this work.
• achievable timeline
  2-year recharter
• minimum viable product
  Outputs: reports, workshops, and coordination outcomes
• does it raise architectural issues that need to be addressed before it can succeed?
  Not yet.

Special considerations?

• risk factors
  The main challenge is sustained engagement and maintaining relevance amid rapid AI evolution
• incentives
  The growing influence of AI on the Web presents an opportunity to explore its long-term impact, identify future standardization paths, and help shape the AI-related Web technologies.
• openness, decentralization
  All deliverables will be public; group discussions will be conducted on GitHub and public mailing lists.

Procedural

• Advance notice to AC when team feel a WG charter is likely to be developed, we send "advance notice" to the AC and public-new-work.
• Horizontal review: in this phase, team should request review of charter-in-development from a11y, i18n, arctec, privsec.

[TzviyaSiegman]

Sustainability is not a horizontal (yet), but would it mitigate some potential concerns to call out Sustainable Web IG as a coordinating group?

[plehegar]

Advance notice

[michaelchampion]

The current AI/Web situation reminds me a bit of the web itself about the time W3C was founded -- lots of hype about possibilities, lots and lots of money being thrown around, but the best-funded players seeing it as a zero sum game rather than an opportunity to collaborate for the benefit of the larger world. Having observed W3C for a long time and seen both successes and failures, here are some thoughts about how W3C could once again be a catalyst for an "everybody gains" rather than "winner take all" Web / AI ecosystem. In short: build a motivated community, help it identify specific problems that W3C standards / guidelines could help solve, and select chairs and editors from the leaders who emerge in that community.



W3C has built successful IGs in the past, such as the Web and TV Interest Group proposed 15 years ago. That effort really did attract a critical mass of industry players who defined requirements for work at W3C, which was eventually recognized with an Emmy award for standardizing the "full TV experience" on the Web. As I recall the proponents themselves -- with the Team's assistance -- built a community of web and TV industry participants, held a successful workshop that identified specific problems that standardized APIs could solve, and ultimately generated energy to drive that work at W3C.



For a Web and AI IG to have the same kind of impact, proponents should FIRST identify specific problems with AI on and the Web that standards and/or guidelines could plausibly address, build up a community who want to develop them at W3C, and rely on that community to identify chairs and editors for the IG. The Team would support the community that is energizing the IG proposal, not BE the proponents



Specific suggestions:

• Hold the previously proposed Workshop and see who DOES come, what they want to discuss, and what focused proposals for next steps emerge from the discussion.
• If for whatever reason you want to be more "nimble" and move directly to an IG, the Team -- and whichever W3C community members are driving the demand for W3C work in this area -- needs to do the kind of problem definition and issue triage that happened at the Web on TV workshop that resulted in a fairly focused charter. The Team's analyses of AI and the Web and the Agentic Web have lots of useful input to that sort of triage, but that's too exhaustive a list to charter a group whose mission is to prioritize it.
• DO A LOT OF REACHOUT to key players you think would welcome the opportunity to mitigate known problems with the current AI on the web paradigms, or collaborate to incubate new, less problematic approaches. Some of the key AI players -- Google, Microsoft, Apple -- are actively engaged in W3C. Do THEY see opportunities for this proposed IG to build a less zero-sum environment for AI on the Web? And some who have been humbled by their lack of early success in the AI boom -- Intel comes to mind -- seem to have ideas for a better Web and AI ecosystem, do THEY see opportunities to explore those ideas in something like the proposed IG?
• Alternatively, maybe W3C's best play here is not to facilitate BigTech initiatives but to rally those impacted to push back on the excesses of AI on the Web. @TzviyaSiegman mentioned the Sustainability IG and the concerns it focuses on. Clearly many people see the energy requirements to train today's LLM models as a significant problem for humanity... can W3C galvanize a community to effectively advocate for regulatory or voluntary constraints? The Team's analyses mention other adverse impacts the current AI systems are creating for today's web, can W3C effectively organize mitigations?


In conclusion: Although W3C-wide discussion of this topic is urgently needed, don't rush to get this draft charter to the AC. First build an engaged community motivated to do some subset of the work identified in the draft, then the Team can help that community with mechanics of chartering, choosing chairs/editors, and getting AC approval.

[Full disclosure: Claude.ai helped turn the original somewhat gloomy draft of this text into what I hope is a useful and encouraging comment]

[hidde]

Sustainability is not a horizontal (yet), but would it mitigate some potential concerns to call out Sustainable Web IG as a coordinating group?

+1

[plehegar]

"A formal Advisory Committee Review for this charter is expected at the
end of August 2025."