Ecosystem & Governance (EGWG) Monthly NAEU - 2026/09/17

Hello. I thought I was in the wrong meeting or something. Sometimes there were double links. >> Double links. >> Hello. Hello. >> Yeah, I have a new budgy. >> You have a new budgy? >> Yeah. >> Do you always have to give the same names or you give him a different name? This one was named by the pound. >> By the pound >> in the animal shelter. >> The shelter. Okay. What was it name? >> Yeah. >> Her name is Pez. P E Z. >> Oh, Pez dispenser. >> Yeah, like a Pez dispenser. [laughter] Anyway, she's a very nervous budgie. And she doesn't want to go back in her cage. [laughter] Well, I'd be nervous to go back in the cage. Well, [laughter] yeah. [clears throat] Um, so we did not publicize this meeting and we haven't met for for weeks or months because of the break and then the GDC and so, um, I don't know if we're going to get many attendees, but it could be a good planning session because we've got a couple things happening. Sankerstein is working on some governance stuff and John um put John Phillips put some governance stuff as well. So hey Stephen is this recording it just automatically recording >> automatically records unless with the host key you stop the or pause the recording but the transcript will keep going. Okay. Um, let's see here. How's it going, Lynn? >> Very well, thank you. Good morning. >> Thank you, Mayor. You guys ready to join the EU? Well, that remains to be seen. I don't know. Pluses and minuses, I guess. >> Associate member, not a full member, but >> yeah, >> sounds reasonable. >> Regardless of the terminology, it was a pretty cool speech, I have to say, from my side of [laughter] the world. >> Now, now, now the tantrum is is hitting us. [laughter] He our president's so upset he's gonna stop trading full stop with the EU. >> Yeah. >> Yeah. Go for it. >> Cool. [laughter] >> Yeah. Um >> after he demolishes the Kennedy Center [snorts] is justified. I mean put his name on it. So Well, let's look at um let's look at a couple things here. >> But speaking of governance, I know Sam Shan's on the call, so that's cool because I know he's written a lot of stuff. >> Um for those of us who happened to be on the steering committee call yesterday, I thought the >> when it was declared the fastest meeting ever, which was great. Uh it was the after meeting part which was actually much more interesting for me. I think uh whatever Drummond revealed from that meeting yesterday was pretty exciting or interesting I would say. It's fascinating to think that uh had Bernie Sanders on stage followed up by Steve Bannon. So I think that in itself is is a great >> uh photo op if they took it. I'm sure they did. But um I [clears throat] I didn't catch the name of the book Drummond was referring to. If we all if we built it, we all die or something like that, but >> I was quite captivated by it. Yeah, I'm I've not read that. >> And today even what uh King Charles had some meetings >> AI >> in Scotland maybe about uh putting some kind of governance around AI with all the other big leaders there and Microsoft is saying that um there you go. Thanks. Um Microsoft was saying now is the time. So basically we know now what what it's not a bad logic if you want to reverse uh think it or reverse engineer it look we'll build governance around what we know what we're building governance around so I think you know the question is do they they're not going to stop obviously but uh the governance may have could come out a bit earlier especially since open eye now admitted six more breaches to put it lightly >> I mean but There there there was a video that I watched yesterday that that talks about this and there's a group that's actually uh actively uh you know going to release additional ones and and and the scenario for getting rid of humanity is well AI wants more and more resources. So okay uh do we give people more resources or do we grab more resources? So it it's not it's not it's not a deliberate decision. It's just we're going to put ourselves first. Yeah, they I mean just being in Europe and ever let's say the last three years when it's really gone warp speed with AI and and Europe's been beating the drum about governance and regulation and and the US is like no no we're going to regulate ourselves or or let the market regulate and all this kind of stuff. So it's good to see that it's it's they're taking it seriously. But the um the narrative that I hear as well is that all these companies are like we need we don't want open-source AI to to gain any traction and so by doing this it allow them to sort of keep a strangle hold on the development etc. Well, I mean on the other hand, if everything is open source, which if I understand correctly, I mean the next hour will be Wayne Jean's call, which will be great because he always has really good insight into these things. Um I mean the problem with open source is anybody could get open source [laughter and gasps] and and if there's a governance layer or waiting whatever they can go, yeah, I don't think I really want that or here's mine instead. M >> so I don't know how you regulate uh open source by the time you get around to regulating it the the genie or the toothpaste is out of the bottle. I mean it doesn't it's like what was the what was the line uh they gave or or Drummond gave an example of it's like a race a nuclear race who who pushes the button first once you push the button it makes no difference who pushes it second. >> Yeah. Yeah. Yeah. they'll be regulated the same, but I I don't I haven't followed that thread, but that's just the sort of >> unless there's some kind of executable governance, right? I mean, legislation, by the time legislation comes in, we're just like a puff of smoke at that point. Who cares? >> Well, I mean, in Europe, they won't even let them release the models. I mean, so it's [clears throat] I [snorts] I don't know. I don't know. Should we start this this meeting though and we can continue to talk about that because there's some AI stuff and then I'll I'll go ahead and share my screen I think. Gosh, I haven't been on Zoom in a while. So, and sorry I missed the uh steering committee meeting. Where is this? I think it's this one. Okay. Can you see my screen? >> Yes. >> Okay. Okay. Well, let's go ahead and start the meeting. The recording is has um started already. Antirust policy notice. Um no need to read it out loud, but you guys can read it here. Is that legible for you? >> We're all members anyway. >> Yeah. So, we've all done that. Okay. Um Carly has um she's retired as co-chair, but um nice. Yeah, she's she's smiling. Big big smile. Um but thank you for your service, Carly. And um we will continue to forge forward without you and or with you just joining the meetings. So thank you for that. Any other announcements we want to talk about before uh we get into the meat of the meeting? Anybody have anything else to offer? Do you want to talk? >> What's that? >> I just said everything I had to say. I'm done. >> All right. Okay. Do you want to say anything about byite-size trust? Carly, >> you're on mute though. >> Uh there are three task force updates now. [clears throat] So number one, governance architecture. They are putting it as a pause on the calendar but holding the spot. Byte-size trust. I'm in the midst of trying to work on some scripts because trust query protocol is coming up to the 1.0 release and it'll be great to have some scripts associated with that and I'm looking for feedback. I posted one in the trust registry query. I mean it's on the byite-size trust document. Uh and then number three there is a third task force under ecosystem and governance. Uh and that new task force is the reporting task force. >> Yeah, I was going to ask about that. Okay. >> Yeah. So the reporting task force right now um the so this task force chairs will need to uh be ready because reporting is coming up. Reporting for everyone is coming up. We always schedule it now right before I for both the all members meeting slides and the TAC technical advisory committee report. >> So I'm coordinating with Thomas the uh in the reporting working group we came up with like a prompt. So then the different members different different ecosystem working group uh sorry different working groups and task forces can either use the prompt or discuss manually to pull together the content for the TAC. So that should be coming out in an email from Thomas the >> I ask what are you prompting? >> It's a it's a it's a a skill. So it it' be like um there there's a report and you can fill out the notes about the report manually and then send them back to Thomas or >> where is the content coming from that's I I don't understand the user >> that would be it would be from from it depends on which work what the working group uses. So for example for ecosystem and governance we would probably take it from our confluence notes or the meeting transcripts or and or uh if we had anything on GitHub. So because different working groups have their conversations happening in different places. So >> okay well in our in our case right we're not too good at we don't bother with notetaking so it would have to be the transcripts. So now you want to take the transcripts from all the calls >> and >> what put them into Google notebook LM. >> Yeah, I guess and then try with the skill uh like we're we have never we have never tried this before and I was try and I was trying to make it easier for everybody. You can also take from the steering committee meeting minutes [snorts] if there is anything that was discussed there. Use that as another source and uh but there is the manual the human version of it and you can also just talk about >> there's this thing called Rosie as well right? Can't we feed it into Rosie? >> Rosie is mine. [laughter] No. >> So, here's this is the >> the drafts of the the human. So, we've had a we've had a a um a uh a form before. So, if you want to just fill this out here while you sit in a meeting and talk. So, if you want to lead by example for the other groups, right, we should probably try to do that ourselves first and then say this is how we did it. >> Yeah. >> And I think the uh most consistent >> way to do it would just to say download all the transcripts. >> Yeah. >> Um put them I would say notebook LM. uh the only ones that would repeat I I don't even think you would want to inter interfere with let's say steering committee notes in your working group notes I think you should keep it here's my working group >> here's all all the transcripts >> and you could say whatever prompt you you want to come up with whatever you need at least you play with it like that and then you say or you say fill out this report for me >> right I don't think it has no I'm surprised that within Zoom since it does have the AI note takingaking. I would have imagined that it would have this kind of ability somewhere within Zoom. >> Yeah. Yeah, it does. I don't know where the >> where the transcripts are saved. So, the transcripts are saved as what likevtt or something, >> right? I mean, but you have to go and download every every VTT individually and do it. I'm saying since Zoom already has the note takingaking capability and since Zoom already has the transcripts, it seems like Zoom would be the one to best say if you if there is a way and I don't I don't have a Zoom account myself. So, but LFDT would certainly know this. Most of the work that you need done is probably available somehow through Zoom and we just don't know the right door to open. >> That's step two. So, we're trying to do it to be more automated, but that's a that requires a pipeline and and access to Zoom accounts and stuff like that. So absolutely that would be uh so I'm looking at so if you if you look at your meeting dashboard [snorts] uh um let me let me uh >> so >> if you if you go to the meeting dashboard >> yeah exactly in past meetings >> right you can look at the recording download the VTT files and down the AI summary as well. >> Okay, >> if you want to download the summary, >> but I I in my opinion, I would take the whole transcript, not the AI summary because the summary is somewhat >> Yeah. >> So once you're letting it consume all that content, I would say >> if you want to try it manually, it's not hard. I mean, we could do it for a few months. just test it, >> download it, put it into notebook LM >> and then give it the form that you want field and say fill out the form. >> I there's a there is a there is a prompt that we used and I shared that in there as well. So it's the second link in of the docs. So that one had it was more than just filling out the form and that came from the work that we were doing the last time when we wrote the TAC report Drummond and me and then shoot what's her name on the TAC she wrote that so uh um so so it's a bit it has a bit more than just here are the headings fill it out >> okay so this will be uh all this will be worked through in the task force when is the task force Uh >> uh the task force meet this this this um >> yeah uh uh we could so in the task force then next week it it's right now it's still meeting once a week we could meet and and test that out with uh ecosystem >> when when is it when on the calendar is it >> yeah uh it is 12 till 100 p.m. on Wednesday >> Eastern >> Eastern time. >> On Wednesday, what' you say? Wednesdays. >> Wednesdays. Yeah. >> Okay. >> And then the other thing is that we are uh trying to work with other Linux Foundation members to come up with an more automated pipeline. >> Right. [snorts] The invitation. Um, so Mackey, I think you go to Do you have the link? Does everybody have the link to these um your individual dashboard? No, actually it's not that one. It's this one. So, you just go to the Linux LF the calendar. >> Yeah, there there I am. >> So, reporting. And does it have a send invitation? >> And you can let me see. Can you do that >> either that? Yeah. I don't know how to add another invitation. >> Okay. No, there >> Yeah. Or are those past meetings? So maybe go to the future meeting >> and maybe they're the >> There we are. public calendar or or you're a signed in one. >> What's that? >> Are you in the public calendar or >> I think I'm in the public one? >> I think I'm in the public one. >> Well, that should still allow you to add a meeting. >> So, let's go to Wednesday right there. Yeah. And then register. >> Register. Yeah. >> You go to a future meeting, then you register. And then um I think at that point it's in all of your all of your all the meetings will then populate to your calendar I believe to your personal calendar but I'm not sure. >> Yeah that's that that that's correct. I I did that initially. Yeah. >> Yeah. Okay. Um >> so a question to Carly. What what is what technology is the skill assuming? uh Frontier AI model that you can publicly access and I've made a free account on one of them on chat GPT and gave it the skill and gave it some of the files and but Google LLM would probably also notebook LM sorry would also probably work but but it it can also be done manually. the [snorts] group can just sit and try and remember what they did. It was just that the AI skill was to try and smooth it out, make it a bit easier if people have >> We also want to be consistent, right, across the groups. Everyone should be using the same inputs and and hopefully the same prompt to get the same so the results can coales at least. >> Yeah. I I I mean, we're going to keep trying. This is a continually improving system because up until now it's been Drummond and me. >> Yeah. Okay. >> Trying to remember. >> Okay. So that's >> if you ask Linux Foundation if there is >> some feature. >> We're working on that. >> Join the Wednesday task force. Um and and you that's where you're going to be working through all these questions, right? >> Uh technically I mean technically ecosystem should do it on their own. Yeah, I know. >> But but I'm gonna so so the next meeting then uh we're going to discuss the proposal from AifaQ about running our own rag running our own rag and what it would take to run that at Linux Foundation and also then we can test uh um ecosystem at the same time. [clears throat] >> Okay. But at the moment we don't Yeah. So we don't take notes. We're just relying on transcripts. And so however we feed those transcripts into I mean it shouldn't be difficult. You just say listen this is what I want the output. Take all the take all of the transcripts and and fill it in. >> Yeah. >> So okay. All right. Thank you. So any other task force updates? Oh, and the official name of the task force is reporting tools task force. >> All right. So, just just a note, I I took the action item to um ask for the changes for both the uh governance architecture group and the data modeling group in terms of quote suspending them. So, I'll be feeding the details. Uh the intent would be to have notes hosted on Discord specifically on those channels uh and also quote suspend the Discord channels. I need to have some technical discussions with people with what that means. Um but that's the the intent. So you know they're not dead. They're there. It's not that there's so much going to be suspended is that they're both going to move from regular to on demand. >> Which which task force are these? the data modeling and the uh architectural uh ecosystem architecture task force under John Phillips and company >> uh the governance on this one. >> Yes. >> Okay. So data modeling. So um so there so both of these are pausing. Right. >> Well, they're not pausing. They're just going from regularly scheduled show up and you'll have something to on demand. >> Oh, >> right. So we're not getting rid of the time slot. We're not going to drop them. We're just going to say these things are not, you know, don't don't show up assuming a meeting's happening. Uh they're on demand. So expect an explicit posting in Discord about an upcoming uh schedule. >> Okay. >> Or upcoming meeting. >> Does that sound reasonable? >> Yep. Yeah. Got it. >> We may eventually move that way if we don't uh get our act together. So let's uh see what we got here. Um, okay. So, let's go ahead. So, this is uh, you know, we've been talking for a couple of months about the direction and where we want to go and the feedback that we've gotten is that we should really, uh, it's time to look at the governance metal model which is, I think, five years old. Um, it doesn't account for things like AI and agents. It's uh as these trust frameworks are becoming a little bit more mature and the Glyfe uh framework for example have or the VLEI governance framework having gone through a couple of iterations it's it's a good time to to assess and so that has been the consensus of the working group and people who who've um made their thoughts known that they'd like to work on this and so the question is how do we do that And um so we we've got here in this in this um uh little box some links to the um the VLEI governance framework. We've got this um document here that um that John Phillips has [snorts] um asked various AI Gemini chat GPT Gro Claude I don't know what system is Grock and Claude um he's asked them to to assess the governance framework um and you see the prompt here does have that. I'll put that in the in the meeting link here. Um so this is something we can review um on this call. Let me go back to the thing. Um and then we've got um Sankhan who has put this together I think. Is that right? Sankhan are you the author? Is Sashan still with us? >> No, not that I can see. >> Okay. So Sanka put this together. Governance, authority, and assurance meta model. So we could go through this and see if there's things here that can then be um fed into the updated meta model. And this has a you know GitHub repository where you can see all of these things if you want to look at that version. So um so that's that's the the focus. Now the question again is how do we do it? You know what's the approach to going through this? Is it just a really methodical take a look at the meta model? Go through the table of contents. Go through piece by piece. What would people suggest? And I guess I should pull up the meta model, right? I think based on discussions we've had with Carla is that it would be great. She I believe she has a bunch of notes >> on what changes are required because if we just go through it from beginning to end, we're probably going to go, "Yeah, that's good. Yeah, I like that." So since she actually has real world experience uh we should start with her feedback as uh maybe new requirements and then go and update the model afterwards. I don't personally if we just review it I don't have real world experience with it. I couldn't say yeah this is no longer valid or we didn't we skimped on this. Okay. So that's so that's a sort of a glyled approach. Any other suggestions or approaches? >> Haven't we found anybody else using it? Scott, >> I've asked a mind. >> I've asked uh the Bhutan NDI folks and they said, "Yeah, yeah, we're using it." It's like, "Wow, you know, how exactly?" and they're supposed to get back to me, but I don't think I don't think they're using it word for word. I think they were inspired uh by its in spirit. >> Yeah. >> Right. Scott said he's used it, right? So, he must have his own notes as to what updates he'd want to make anyway. I mean, that's a >> to me would be the straight from the horse's mouth. >> So, here I mean, let's look at this. This first one would if we said that the primary document all right it's a homepage for the governance framework it must have a did is that I mean is that something that is you know still relevant does it does it have to be a did or do we want to update it to uh to to vid which is the higher class >> any do any docs have a did >> or vid do any of the docs any of the specifications delivered any deliverables do any of actually have a real life did. >> Well, they should. >> At best, they have a Shaw [clears throat] from GitHub, >> right? But I don't know if they have a uh triangular did. >> So, I think that's >> that was probably um nice to have, not a must. That was wish that was inspirational. I think when it was originally written seven years ago, that was the inspiration that Oh, yeah. every document ever did. >> Okay. But so that Yeah. So that's I guess the the the conversation we should have as a group is is that something that should be updated to say that it should have a so the trust spanning protocol I think and maybe is even the trust registry query protocol have introduced a concept of a vid right so it's a verifiable ident identifier as a sort of an overarching class and within the vids you've got did you've got >> aids the autonomic identifiers >> and even X509's are verifiable, but they're not decentralized. So, should we update this document to say that um the governance framework must have a a vid? >> Does GLE give it a >> I don't know. >> Has gly given it a did. >> Yeah. So, I guess how do we do Yeah. Do we want to actually just um I mean because these are good questions that we should go back to and somebody should take notes. Um maybe I should take notes. Um does glide give the uh the lei governance framework a did okay that's a question. Okay. Um, and then let's do this with um should we update must have a bid to vid. Okay. Um yeah. So I mean these are the kind of things I mean we we kind of have to either ask an AI to go through it like like John did and say you know what's missing where are the gaps and that that will be a little bit more comprehensive or we just go through it methodically and say you know what are our opinions or we wait for GLE to lead. >> So I know Stephen you're voting for Gly. Anybody else? Well, I mean to some extent huge numbers of documents have been thrown around on this discussion. We need a catalog. Um uh I I've tried to capture them as we're going through here, but I don't have that catalog. Um uh using gly as as probably somebody who's had to think this through commercially as a lead would be found is actually requirement and not. >> I suspect that uh the combination of two is going to be very fruitful. Carlos, nothing if if not thorough. Um, so you know what's that look like going forward? But you know, we need to have a repository of all the information that we're dealing with here. >> So, >> okay. So, we need to create a document with all of these um references such as this one that Sankashan has put out, this one that John Phillips has done. Okay. >> Yeah, that that that would certainly be a start. >> Yeah. Okay. What else? I uh I asked the a Glyfe AI they have a they have their own rag >> it looks like and they say that yes they use decentralized identifiers but they haven't said that they used it for their darance document but then they switched from DIDs to car's aids autonomic identifiers >> okay So, I'm gonna keep looking. >> Okay. Um, okay. We're all becoming part of a carry-on movie. I can carry onward. >> Everything is carry. It's carry all the way down. >> Yeah. Okay. Anything else? [snorts] What about I mean having a web version of this? It would it shouldn't take too much to to turn this PDF into like a website that you can navigate. >> Does somebody know how to turn it into a GitHub pages which would be we could then put it up on GitHub and have it accessible as a GitHub pages on uh the trust over IP GitHub. to GitHub what pages >> certainly tools available but maybe even um Google Docs does it if you could import the PDF and then say export to markdown >> no >> markdown is native GitHub I don't know how to do the pages certainly does >> as [clears throat] we saw in the link um but yeah I think just starting off with uh you know markdown is there's a markdown editor You could just do it directly in GitHub. We don't have to make it more complicated than that. >> I mean, actually, we should, if we're going to work and dissect this, we should probably turn this into a a Google doc or something so people can make comments and we can >> Yep. Good idea. >> Yeah. Having been through this recently on on another project, um, generally speaking, it's first first pass for general consumption is is Google Doc. >> Yeah. Um, and then for version control, it then moves into GitHub and it becomes Markdown, >> right? >> So, so I think that's that that's that seems to be where where people are trending. So, hey, look at this. So Carrie the sorry the VLEI governance framework there there's a little bit of funniness about the language because in in 1.1 about the did >> Mhm. >> we said that the organization >> needs a did >> who we did. >> Okay. No no okay the primary document. Okay. So the primary go document for the governance framework and it must have a DID that serves on the identifier of the entire governance framework. Okay. So the primary document needs a DID and when you download the VLEI governance framework version 4.0 it's called primary document >> and it has a document name and it has a document did URL which is did Harry blah [snorts] blah blah. >> Okay. Okay. So good. That's that's so far they've got two for two. >> Yeah. >> Um and does it have authoritative references to all other documents in the governance? >> I'm looking because it's all on the web page, but I don't know if it it doesn't reference it from within this document. Soon said, if there's a planned activity to update the meta model, it might be ideal to create a list of inputs that could be examined to determine whether an update is a minor change or a major release. At this point, I don't think we have that input requirements in place yet. So inputs, a list of inputs. What's an input? What's an example of an input? >> The VI governance that I put a link into that we can use as reference. An example of an input could be why do you want to >> change or update the meta model itself. Like if somebody comes and says that meta model doesn't cover agentic systems and it doesn't cover certain specific aspects of agentic systems. >> Yeah. Yeah. So you have a well scoped gap that can be evaluated as does it does the meta model really need to include agentic systems or if it does how does it cover that and so on and so forth. >> Okay. >> So >> yeah sounds like a GitHub issue to me. Uh yes, >> it it does work well if it's a GitHub issue, but uh and I see a quite a few groups in trust over IP adopting that path. I'm not sure if this working group is keen to go down that route. >> Not yet. a list of requirements or inputs requirements that will um help determine um whether or how the the governance framework might updated. Okay, so we've got Yeah, I mean I think that's the big one is AI and Agentic AI systems. Um, yeah. So, maybe we can Yeah, Mackie, that's what we're trying to get is we're trying to get Carla because I think Carla's been taking notes over she they have to update it like once a year or something. So, she's been taking notes. Um, should we start a document now? >> So, we can, you know, put this in a in a document that then is distributed and people can contribute to. >> Yeah. Are you going to do it in Google Docs? Because my problem is now nobody every time I try and share something from the trust over IP Google Docs drive nobody has access even people who should have access >> and and uh uh uh um the uh we we could put it in GitHub markdown and I'm a speaking as person who hated GitHub [laughter] at the beginning but eventually I learned it doesn't have to be trusted of IP Google I mean doc does Yeah. Okay. >> Um, >> discovery is a bit harder then. >> Well, then somebody's going to have to own it. So, if we say ecosystem governance working group, let's see if we can do here. No. >> Yeah. >> Here we go. It's called um governance >> data model V2 >> governance meta model um V2 planning or whatever. Okay. So let's call share save it. Let's just copy a link and let's see who gets it. Who can access that? >> I can access it. >> Everybody. Okay. All right. So, this is in this is within the shared drive ecosystem governance um working group folder. And it's um so so okay everybody's got it. So let's work on this document. And uh where'd it go? Right there. And um we'll do the list of requirements why we want to do this. will have a reference of a catalog of different references for Sankers's work and for John Phillips document and to the Golia VLEI governance framework etc. So we'll start just working from this document to um to kind of work our way through this whole process. Everybody happy with that >> about the alternative is based on a metal model but not not exact. So I think Sakshan I think so Gly have I think the the Glyfe version is practically like mirror version of the meta model but it's in all of the control documents where things just become um tailored to the specific environment and the use case. So I think if we look at the um maybe we can do that maybe we can run the GLY framework the BI framework through AI and give it the meta model and say does this match up? Is there anything that doesn't match up? I mean I don't know what that will tell us but >> I think that would be problematic. >> Why? Because the meta model started off maybe like 14 pages and gly must be hundreds if not thousands of pages. >> Yeah. Engineer to the 14 pages and see how far it's diff. It's going to be >> thousands of pages of diffs. >> Um well you can give us some guidance monkey. Yeah, actually I think it's it's a good idea if we can start envisioning how that would look like from our own perspective and then I think you know meeting with Carla could you know like we can then go through it with Carla and see what's missing uh because you know like really at the end of the day I I think she should be able to gather some feedback from the community who who used it already. Uh but I think from our own u you know like proposal like we can propose something and then connect the dots with with her. Um yeah at least you know we can find gaps uh uh maybe you know we can envision for example that it needs to include agentic as you know like uh Sankrashan just mentioned >> you know like any other uh dimensions you know we are we see that would be a requirement for the next version um and then you know we can work backwards uh or even adjust uh based on her feedback um yeah >> yeah okay so we'll get Carla we'll we'll We'll um send this document once we get it started. Um this one, not that one, this one. And get uh so we'll do I'll do an introduction and start making this document a place where everybody goes to figure out what's happening. And uh we'll insist that we we really need Carla's help to get us going on this journey. Yep. Yep. Yep. Um Carly, >> I think we're going to have to eventually have this in spec up T. >> Okay. I mean, we want to get it to a point where it's >> so >> so then it would make sense to actually get the the meta model either word doc or get a a word version or a Google doc version and then >> I mean >> and then export in a smart well then figure out how to get it into specup t. It might be I maybe I will look into starting at specup T because it might be a lot of work to retroactively fit it. We'll see. >> Okay. So calling to expect spec um sp Okay. Do we want to take a look at this um at this? Where' it go? Where did it go? This one. No. No. Where's John's? Should make that bigger. So these are based on this prompt. I want you to act as a PhD level research assistant and research the state-of-the-art of governance of AI. So this comes from the perspective wow is this today done today um that that they believe that it is missing uh it's it's not perfectly relevant for uh AI at the moment and so they want to figure out where the gaps are. Is the TOIP governance framework suitable for AI governance as is? If not, what needs to be done to enhance the meta model to address AI governance? So this is the early returns telemetry. Who can tell me what telemetry means? >> Basic. It's [clears throat] basically the equivalent of airline, you know, an aircraft telemetry, right? It's basically it's a diagnostic uh stream. >> Okay. Byte-size trust. Carly telemetry. So replace static risk assessment documents with a telemetry. An observability control document. I >> think you say telemetry. >> Telemetry. >> Not a telemetry. No. >> With a with a telemetry and observability. Well, yeah. >> That's a Canadian pronunciation. We say tele. [laughter] you in there somewhere. >> I I'm still don't understand what it is. Telemetry. >> It's actually try to tell me. >> So replace static. So at the moment the the risk Yeah, it's probably right. The governance meta model says all right identify the risks and then construct a um a framework that addresses those mitigates those. So the government meta model that we have right now takes the uh risks and puts them in some sort of a what is could be described as a register and then through various uh processes allows the evaluation of the risks in terms of how they would pan out when the governance framework is adopted by any deployment. I think what uh John's AI adventure is suggesting is that this one-time activity will end up being fairly fluid and thus no longer it's static and end up being versioned so that at any point you are able to run audits based on a particular version and figure out what your risk profile is and how is the government's framework managing to handle the risk. Yeah. So this is the the I mean just the general trend in cyber security and and others continuous authentication continuous um um well observability. Yeah. Maki. >> Yeah. actually you know I I just wanted to mention that to move away from that the static uh you know note version um that's the reason why we wanted to have some tools to enable uh implementation like the risk and principles document we started with so I think this will be very fluid like later on if anyone would like for example to adjust or add to the risks and and the uh principles that could be done so I think um this could be more into the implementation tools like enabling the governance metadata document uh basically Um uh but you know like I just had a a comment in terms of the second one uh the telemetry and observability when it comes to mandate carry anchored continuous machine controlled I think it will be safer to have machine readable as opposed to machine controlled you know machine controlled could be like aentic which which could you know raise a lot of threats like you can use agentic but you need to have some sort of you know like um you know guard rails or or you know some some um uh methods that you can um basically have uh human enablement in the process. So I think will be good to have it as machine uh readable. Um Gary anchored I do agree with this. U I think things are are shifting towards Gary. U we've discussed this before as well. >> So how would that work if you've got if you've done a risk assessment? You're like okay we know industry we know our use case. you know, um the data we're playing with whatever we've got our risk assessment and if you move that to an ongoing sort of continuous risk assessment, how does that get fed back into the governance framework and updated >> or the governance document should be? Yeah, I think it should be as broad as possible like without really becoming like a policy kind of uh document. I think it should be like a high level guidance and then you know the tools could could enable us to fluidly you know uh implement uh also the tools can can help us you know like um update rapidly like uh like moving away from a static document. So I think the next version should be like very uh I would say u fluid like very broad uh document broader than the first uh version. >> Okay. I mean I think we should it would all be good it would be good for everybody to if we just reread this because I think it is fairly generic and all the real >> it is designed to be generic. >> Yeah. So for instance, if I go back to the question you had, governance framework at a meta model level will essentially just mention that >> you will need to have a risk register of some sort that's likely machine readable and continuously updated and managed and controlled. It is not going to give an entire let's say a vocabulary of risks to populate a risk register or it could give it as an example in an appendix of sort but the main uh declaration would be that you would need that you require that artifact to have be present. >> Yeah. >> It will not go deeper into what the shape and form of that artifact would be. >> Right. Yeah. Okay. Yeah. So that just missing. Huh. So, so just a thought here I mean [clears throat] what I think is missing is you know uh static risk assessment documents with operational tone watcher it's the decisions that are being made you know under the opaces of a specification right so you need to have observability of decisions being made using the specification right if you design a process for you know flights for aircraft >> you know there's There's policies about how you fly, how you, [snorts] you know, how you follow a route, how do you interact at at way points with air traffic control. Um, and the telemetry is all about the actual operations. It's it's the verification that the process is being followed. >> Okay. >> Um, >> so telemetry about the documents is uh you know the specifications is nothing more than version control and version handling management. Okay. So, yeah, that's going to have to those are the kind of things are going to have to go into people language. I don't understand any of this. >> I I don't even know what like so telemetry is like real time ongoing collection and transmission of the data. And how on earth would >> well it's it it's not necessarily the data. In this case, we're talking about documents. It's about decisions that are being made about the implementation of the specifications. >> But they're saying replace risk assessment documents >> with a telemetry >> an observability controlled document. Like >> is that is that just slop? That's like implementation. >> Yeah, I think to me that it's it to me it says you were going to work on git control documents and if you make changes to those git control documents then there should be a notification. So I'm not I haven't looked through this document that John had shared uh but I think given what the prompt is it is attempting to reshape the existing meta model framework and trying to highlight if the meta existing version that we have were to be expanded to cover agentic systems what are the capability gaps that it should also factor. in so telemetry, observability and all of that are capability gaps. Now I think that one of the things that I should emphasize is we want we would like to begin by considering that we have this existing meta model version and then create a requirements document that is gathering from >> as many inputs sources as possible and then evaluate in terms of how the next version would emerge. So the agentic document that John has is not is very focused on one particular problem to be solved. >> Ensure that a metal model meets the agentic world. >> Yeah. Yeah. Yeah. So I mean that that's I guess that's the point is that yeah we we need to there's you know no shortcuts. we need to go through these these documents and this is very specific to the prompt of you know is it suitable for for AI um and where are the gaps but that's just one aspect and if we um so we've got here in the notes to um begin a list of inputs requirements that will help determine whether and how the government's framework might be updated. So what's missing? What are we trying to get out of an update? What are we trying to address? um etc. So I'll try to write that into the um into the introduction. So I'll try to put this here. I'll get a starter text here. Everybody contribute so that we all understand what we're doing and why we're doing it. And then once we're happy with that intro, then we can sort of go off and and look at all these different um um references and start working. But we've got a minute and I just want to make sure everybody knows two weeks from now Verona is going to come and present and they're going to present what they've got. They've been working on it for a long time. It's open um source I believe open standards for sure. They're using all of the standards that we're familiar with trust trust registry query protocol. They probably they might have some TSP in there. They got DIS W3C uh bur fiber credentials etc. Um, so they'll present that, but then they're going to get to the point after the presentation where they're saying, "We've got all this. What we're kind of missing is a nice um, governance framework." And so that will lead to a discussion where we can um, exchange with them about how that might look. Carly, >> so if you share with me the ad and the info, then I have permissions to add it to Linux Foundation decentralized trust LinkedIn. Okay. Okay. Yeah. I've already sent them the link so they can fill in the information and I hope to get that by tomorrow. >> Yeah. Okay. So then make the picture and the and the info and then I can post it I will I can post it up and if somebody then wants permission to post on Linux Foundation >> LinkedIn >> uh yeah or or you can submit to get your own permission to do that. >> Okay. But so this is with Fabris uh who has been part of the trust registry protocol and we we you know so he's part of trust over IP and they've been working on this for a long time. They've they've built some you know interesting stuff and they've got some interesting use cases but their um whole um point after the presentation is to talk about you know how can infrastructure like this be governed. So um please join in if you can. >> Yeah. Yeah. Let's get that ad out because uh trust registry query protocol >> was looking for people who are working with trust registries. >> Yeah. Okay. >> Okay. Good. Thank you. >> All right. Thanks everybody. >> Bye. >> We'll call it a day. All right. Thank you. >> Thank you. All right. Cheers. Bye. Bye.