Meeting minutes
<koalie> aloha!
Hidde: this breakout session is about the open web, a follow up of previous discussions at TPAC and IETF, with sessions chaired by Mark Nottingham about the problems the open Web is facing
<cpn> Notes from those previous sessions: https://
Hidde: I want to focus the session on what kind of things we can actually do to mitigate what's happening
… I want to move beyond talking to concrete action plans
… the fact that the Web is open is something of us love about the Web, and likely a big part of what has made it so successful
… but the Web is being hammered by traffic to train LLMs, with lots of bots traffic increasing serving costs
… that content is also being used for training, which is then exploited without backlinks to the original content
… e.g. if you're a news publisher, your content might be accessed via an LLM without ever getting a visit on your web site
… we're seeing this on our Dutch government web sites
… Wikipedia has also reported a hit on their budget because of that
… Web sites can introduce paywalls, login walls, move off the Web
… all of these affect the openness of the Web
… the session description has links to blog posts which suggestions
… e.g. supporting organizations that are pushing back (e.g. EFF, Mozilla) - maybe W3C is one of these too
… regulation could help
… banners for funding Web sites, à la wikipedia
… focusing on publishing platforms that are easy to leave because data can easily be exported
… in more general, enable more social primitives that enable to distribute content
… re-take ownership of our tools
… using ATProto as a way to manage contextual consent
… does this require us to rethink of how we develop standards for the Web?
… e.G. current work on augmenting robots.txt to manage traffic from AI crawlers
… What can we do in W3C to keep the Web open and address these challenges?
koalie: one path for W3C to fight is through statements where we express the voice of the membership to save the open Web
cpn: I and my organization care very much about this topic
cpn: I'm active in the robots.txt work; there has been a huge value extraction from open web publishers towards a small number of well-funded AI companies, leading to a new form of concentration of influence and centralization
… the way we're looking at this is partly that we want to see a more fair value of exchange, which is not happening at the moment, as shown by the huge drop off of traffic to web site mediated by AI systems
… some publishing companies see this in financial terms - either through a TBD protocol based on scraping/collection/usage that creates value that flows back to content creators
… there are various efforts in that space; e.G. the BBC is involved in project SPUR that would want to support a standardized solution for this
… also relevant the work of describing what you want to allow (in IETF), getting visibility on who's crawling
… also facilitating gathering of analytics that are needed to improve content creation but aren't available through LLM intermediation
… and for commercial companies, remuneration when the content is being used
… right now this is approached through legal actions and licensing deals, but that's only available to large publishers, which doesn't create an equitable system
… right now, the open Web has allowed this via advertising
… if AI platforms become the new entry point to Web content, they would concentrate all advertising revenues
… Without incentives to create and publish content, the web loses its sustainability
<Zakim> amy, you wanted to express interest and ask what W3C as a forum for discussions would be like
amy: hidde, you asked if W3C is a good place to support this discussion - what would be needed for W3C to be such a place?
… I've seen a lot of thoughtful discussions happening on these topics in W3C
… should we look into an event like a workshop to provide a forum for our members and the broader community?
Igarashi: we should identify operations and actions to move this discussion forward
… right now, we're missing important stakeholders (e.g. big AI companies)
… so far our attempt at getting them involved in a workshop have not been successful
hidde: a workshop would be an interesting idea to explore
Dingwei: AI is a new context for the open Web - it's not just about saving the open Web or the Web itself
… we're moving to a new Web - the agentic Web; we should make sure it's an open agentic Web
<Igarashi8> How to invoke Ai tech ompanies are one of issues to to be addressed by W3C.
Dingwei: it's a good time to work on these techs via WebMCP, but it's a bit defensive - we should be more proactive in seeing how to serve the needs for agents
… if we don't it well, agents may move to a closed ecosystem supported by proprietary technologies
hidde: what would be the properties of an open agentic Web?
Dingwei: built on top of open web technologies (vs proprietary ones)
… based on open web principles vs walled gardens for distributing content to agents
… W3C is well positioned to make this possible, but if we don't go fast enough, a proprietary ecosystem will take over
chunming: open is usually related to internet - you just need to follow internet protocols to be part of the internet
… a key point for the open web is that all the service providers are free to set up their web apps and bring it to the Web space where users can discover it
… in W3C's history, we've looked at <something> Web a number of times
… Web services wanted to make a platform to publish and discover services automatically
… the agentic Web can be interpreted in multiple ways:
… - the Web as an API to surface resources, functionalities - compared to the Service Web, the agentic Web would allow anyone (restaurant, university, any entity) to build their own agent
… making it possible for agents to interact with each other will need protocols - in which W3C may have a role to play to develop open standards for
kenneth: the open web is currently getting more closed to humans because servers are trying to protect themselves against crawlers
… captcha, JS-based effort checks - they create barriers to people with disabilities for instance
… these scrapers ignore robots.txt and take adversarial approaches
… getting AI companies in W3C only make sense if they're ready to engage in good faith
… I think we need to support humans before we focus on supporting agents
Igarashi: the challenges to the OPen Web, e.g. walled gardens, pre-exist AI technologies, although they've escalated it
… it remains true that anyone can publish on the Web, but we've seen centralization towards services run by big players because they grow popular
… social media also has played a role of intermediation
… the agentic Web makes this more complicated because their algorithms are opaque to their own operator
… I'm not sure how successful how W3C has been in fighting walled gardens in the past
… not sure there is much W3C can do
Hidde: how can we ensure a successful collaboration with AI companies if we get them around the table
Igarashi: so far we've not been able to attract them at the table
… maybe we need a different way to approach them - e.G. target the CEO level
… or via government agencies
<Zakim> dom, you wanted to react to Igarashi
dom: you're right the workshop from a few months ago did not succeed in creating momentum
… initial focus probaably too ambitious
… the message we got, I think, was that AI agents was so fast moving
… that people would likely not be interested in having conversation
… the topics we see emerging is precisely 'how does web content relate to agents'
… how the sets of products relate to the web ecosystem
… one focus is ecommerce
… but the title of the workshop right now factors this in
… I wish I knew what the right solution to save the open web [is]
… my personal sense: the only impact of W3C is getting people around the table
… number one step
… bring value from our community, our process, our standards
… let people come to the table
… give them a sense of the tradeoffs
… if the web died, they would suffer
… if the quality degraded significantly they would suffer
… this open ecosystem that W3C tries to animate, that's how I see us acting
ota: there is different point of views about the open web: content/service providers, consumer
… for a consumer, the open web is not always comfortable, walled gardens can provide a safe harbor
… transparency and choice are keys to enabling that sentiment of safe harbor
… W3C values the perspective of end users - we should look what's need to support that level of transparency and choice
plh: the open web is a system to share information
… are AI agents part of the "everyone" we're trying to support?
… the cost of infrastrcture to run a Web server is increasing
… because of the cost to protecting the server against LLMs crawlers
… e.g. paying cloudflare - who is now tasked with defining what a human is
… the perimeter of content providers hasn't changed
… the business model of publishers has had to shift a lot along the history of the Web - e.g. leading to the early need (and failure) of micropayments
… for user publishers (e.g. via social media), the need to provide decentralization is served by our work on LWS
… the definition of a user is evolving though - with agents and bots becoming part of that scope
… if AI Agents are more convenient than human-driven interfaces, they're going to be used
… maybe one way to reduce the infrastructure cost is to set up a push-model over the current pull-every-second approach
dand: ... AI is only exercebating issues we already had as previously discussed
… we need to find the value balance
… why do people publish in the open?
… if it's for money, they'll want money back
… if its for fame, they'll want provenance/attribution
… we should address these issues based on the underlying incentives
… When I use summarization from DuckDuckGo, they give attribution
… AI companies act on behalf of users who love these products
… we could have different publication paths - one for users, one for agents
… with proper tools to manage this different paths
… I think we should move way from characterizing this as "saving the web"
<Igarashi> +1
dand: people will keep using the Web, enterprises will keep using the Web
<ota> +1 publisher's incentive design
<chunming> +1 we are not saving the web, but lead the web to its full potential
dand: in many ways, I would want less of my data shared publicly which is creating huge privacy risks (see "my life" scraper)
tink: I don't think the Web is getting more closed, what's changing is the point at which users are interacting with content
… printed newspaper got a lot of damage to their printed business when they started publishing their content
… search engines replaced web sites catalogues, bringing a different business model
… browsers already provide a trusted abstraction of the original content
… maybe we need to look at it about how to make AI agents a trusted abstraction to interact with our content
… including making it a fair exchange between AI providers and content providers
… W3C's role may be to serve as a bridge, not just to AI providers and our current publishers, but well beyond - bloggers, music creators, social video creators
… this would make this also make W3C more compelling to AI companies
<plh> [I would note that the AI companies are (finally) coming to the SDO tables]
<amy> +1 to Leonie, W3C as conduit to attract the right voices
Naomi: W3C standard technologies are vital as a global common language to safeguard the Web ecosystem
<Igarashi> Let’s focus less on description and more on decisions in the breakouts of the AC meeting. Any suggeted action to the W3C?
Naomi: W3C promotes the Open Web, and a number of our members already provide AI technologies
<amy> +1 to W3C as a bridge
Naomi: the W3C Web & AI IG can serve as a place for these discussions
… W3C needs to act as a bridge
<plh> igarashi, the ecommerce workshop we're organizing should help in bringing the AI companies in
<koalie> [I removed myself from the queue given the time crunch and I spoke before, but I want to say back to Dan Druta: AI seems being forced onto users who might choose otherwise if given a choice. Also, the web is being poisoned, so yes we need to save the web]
<amy> thanks Hidde! and all for this great discussion. I hope it can continue more
<plh> Igarashi, if we're going to take actions, we need them at the table
<koalie> [also, I prepared for this session by watching a video interview with Timbl taliing about AI: https://
Hidde: thank you for the input, I expect lots of conversations to continue on this
<koalie> s/talling/talking/
<Igarashi> I wonder that the Team will co-orginize the workshop talk with AI companies with other bodies, i.e.g IETF.
<ota> +1 leoni it's supply chain change issue. we can maintain the web as fare last mile interface for users