World: Building Trust Online | Mateo Sauton at ETHOnline 2026

Hello everyone, this is Mateo. I am part of the TFH team, Tools for Humanity. We are a team Tools for Humanity. We are a team that creates tools for the World project. You may know it as Worldcoin, which is what we used to be called, but we have rebranded.

And so, the idea of ​​this workshop is the idea of ​​this workshop is to briefly tell you about what World is you about what World is , what ideas are in World, introduce you to directions and a few things that we really want you to test and try. Without further ado, let's go. I always want to, I always start presentations with this image. This is the moment when I had a kind of " had a kind of " aha moment." I saw Peter Steinberger, the creator of Cloudbot, say this exact phrase on Lex Friedman's podcast .

I watched as my agent happily pressed the "I'm not a robot" button. And that moment for me was—damn, we've reached a point where we really need a way to distinguish and know that someone is a person online, right? And that's what World does. This made us think that trust on the internet is somewhat eroded, right? Sorry, over 50% of traffic now is bots.

Organizations have encountered deepfakes, a lot of deepfakes, over the past year. You may know this famous case of a deepfake scam involving a CFO in the United Arab Emirates that cost them $25 million. These are just random facts that once again made me and the whole company realize how important this is, right? The need for an AI- right? The need for an AI- proof way to distinguish between a real person and a bot in digital interactions, right?

So, World is a network of real people, created to ensure that everyone benefits in the age of artificial intelligence. We have 39 million active users in over 160 countries, and it's a company co-founded by Sam Altman. The other co-founder is Alex Blania. And we have this amazing product, one of our amazing products amazing products , but our main product is World ID. World ID is essentially a high-assurance, anti-bot authentication method that is a privacy-preserving protocol privacy-preserving protocol .

This means that it contains nothing about you as an individual, but it confirms that you are a human and allows you to prove it online without disclosing personal information. This is very interesting because a developer or you can use this as a level of trust to stop, for example, bot abuse on your site or projects through credentials such as proof that you are human, document verification, or selfie verification. I'll talk about it literally in one slide, it's a very interesting and secure architecture that uses zero- disclosure proofs and generates cryptographic proofs on the device instead of transmitting raw data. So, this basically means that if you get your World ID and want to use it, this thing doesn't reveal any information. It simply confirms that a person is holding this device and is participating in this online participating in this online interaction.

One note: I will answer questions at the end of the presentation. As I mentioned, we have three different solutions for World ID. The highest guarantee method is called verification through Orb. Verification via Orb means that you physically approached the Orb—a really cool camera with a bunch of sensors that took a picture of your iris. She generated proof that there is a person behind this phone , and, in fact, you can use this proof online.

Another caveat: Orb immediately deletes the image after generating the proof. And if you don't believe me, you can check it out for yourself. The repository is public, you can view it on GitHub. It is called worldcoin/orb-firmware. You can check the firmware there, and if you want to be sure, you can check it when you go through verification.

The second solution we have is called Official ID, which is actually a set of passports with NFC chips that we support, and if I'm not mistaken, it's 13 different countries right now . Of course, we are expanding this list. So, this means that there is a government or an institution that we all trust that confirms: yes, this is a citizen of this country who has this document. Again , without disclosing personal information personal information ; the only thing we will know is that a citizen of the United States or, in my case, Argentina, is holding an Argentine passport at this moment. And the average guarantee method is a new thing guarantee method is a new thing that we just developed, it's called "selfie- called "selfie- check."

So, essentially, you can think of it as a " " liveness" verification or selfie liveness" verification or selfie verification that many neobanks, banks, and fintechs are now using to make sure there's a human being on the other side of the screen. We have developed our own version that allows you to prove that you are a person online again without visiting the orb and with a lower level of reliability. This, again, is not as reliable as going through an orb and proving you are human. But yes, it uses your device's camera to prove that there is a person on the other side. But there are different ways to prove it, and it's not as accurate as through an orb.

Regarding the tracks, we will hold two different tracks, 3.5 thousand each. There will be three winners in each track. The first three winners in each track. The first is new options for using the agent kit.

This will be a continuity track. We want teams that are already building something for their companies to integrate agent kit. I'll talk about the agent kit in a moment. Integrate agent kit in a new way that will significantly benefit their project. This could be restricting access for agents without human involvement or any other case they want to show us.

This will be extremely interesting. And the other track is selfie check other track is selfie check , which is currently in , which is currently in beta and will probably be available to everyone next week, but you will receive it after this presentation as a gift from the selfie check beta team. Also, again, create products using this verification method, which is not as reliable as orb, but is significantly faster and with less effort than a visit to a physical orb, as it verifies the image verifies the image your phone has and the image you take during the selfie check process. Um... Ideally, for both options, we would like options, we would like you to provide us with a document with feedback on both products.

Again , selfie check is in , selfie check is in beta. Therefore, we are actively looking for people to integrate it and tell us what is wrong with this product and what they like about it. Of course, we want the same for the agent kit . So, let's move on to the selfie check and a detailed explanation of it. So again , this is a low- friction method for World ID.

This is very interesting because it not only confirms that this person is human and is unique in some way, but it also brings back what we call a "civil score." I won't explain it in a very technical way, but I think most people will understand it this way: the civic score way: the civic score is a reflection of how many times a user has wanted to log in to this platform. For example, if I use selfie verification a million times on different devices, my civic score will increase. So, this is the number of times that face is recognized on the platform. If the civil score is very high, it means that I am trying to violate the uniqueness of this platform by trying to log in from different accounts, that is, I am trying to bypass the system, right?

Again, this uses the user's mobile camera to establish that it is a real person; This is a very quick selfie that confirms that there is a real, live person behind the device's camera. You do not need to be verified through Orb. And it's a very simple process without any extra effort. Unlike Orb, which requires a photo of your iris, this focuses on your entire face. The civil score will be available starting next week.

So if you are interested in improving this new feature and testing it, please reach out to us and I will probably whitelist you for early access to the civil score. And now Agent Kit is the human level for agent automation. Again, I explain it to people like this: you know, we have World ID, right? World ID proves that you are a person online. You can delegate your World ID to your agent and allow them to perform tasks for you with this verified World ID.

OK. So, as it is said here, it is autonomy is said here, it is autonomy , confirmed by man. This opens up a new primitive for agent economics, because you can distinguish between an army of bots, unverified by anyone , and an agent that is supported by a human. OK. And it's a very seamless integration.

You need to execute literally one command. It links your agent's wallet to World ID and secure authorized access, and that's it. Here's how you delegate this to your agent, which essentially writes the agent's address into an address book representing all agents verified by people verified by people . and you can use different application options. On-chain application options.

On-chain delegation is one of the ones I just mentioned. Agent commerce: you can turn to Shopify, for example, or any for example, or any other e-commerce you e-commerce you want to integrate. And if this e- commerce has already integrated such a feature, you can provide special discounts for human-verified agents. Another usage could be " be " human in the loop". This is an integration we made with Verscell.

So, essentially, in Verser workflows, you can set up a "person in the loop" when you want a person to confirm their presence before the agent can perform a certain action. And there is one difference between this hackathon and all the others we have held: this one takes place online. So, as I said, Agent Kit only uses verified World ID Orb credentials, and if you are online, you will not be able to verify through Orb. That's why we have the World ID Sandbox app as one of our latest releases. So, essentially, the World ID Sandbox app is the same World ID app, but in sandbox mode.

This is actually a non-working (test) application. It does not create real uniqueness. But it creates your ID and copies everything we have in the production environment but in the testing environment. It has not yet been published in the public domain. You also get beta access to this app.

That's why I'll need your email addresses to provide access through TestFlight and Firebase App Distribution. For this. So, yes, essentially, this app will allow you to unlock and use all the features that World ID or verified credentials provide, without the need for actual verification. And that's something I find really interesting, because I have a lot of friends and people a lot of friends and people I talk to who are not exactly afraid, but skeptical about verification, but they really want to try what we're doing. So, this is our solution for them.

So I think this is a great opportunity for anyone who wants to use World ID and verified credentials. We would appreciate it if you could give us feedback on this. So if you are building on the Agent Kit track, you are strongly encouraged to use the Sandbox app, and it would be great if you could give us feedback on it. A few examples of templates I would like you to see and perhaps develop further for Agent Kit are new verticals for human-verified agents. e.

g. changing different types of access to certain platforms or different pricing policies for them, as the agent is supported by a human. Another example is agent-based workflows that combine, for example, identification and payments in a way that is different from what we know. Again , I mentioned pricing earlier pricing earlier , not price change, but these are ideas for your discussion; I won't give you a ready-made solution, because I do n't have one either. Another idea or example I want to tell you about is a human-supported agent for regulated workflows.

One person is behind many agents that change the logic of the product. Until now, I haven't mentioned I haven't mentioned that you can delegate your World ID to many different agents. OK. So if you want, you can, for example, I've delegated my World ID to three or four different RMS agents, and they perform certain tasks for me. And finally, this is another idea regarding the “agent-agent” or “agent-API” trust flow.

These are again ideas for your discussion and reflection. One example of using Agent Kit was the Shopify integration we did a few months ago. This is also described in the documentation. I want to show you how easy it is for how easy it is for anyone to use Agent Kit. So, we did an integration with Shopify where we gave people caps.

We had 500 caps exclusively for verified people. You could buy this cap, in fact you had to buy it, but only verified people and verified agents supported by people could get this product. This is a very simple integration. You add the plugin, install Agent Kit, Shopify integration, and then ask your agent or Claude, or whoever, " whoever, " help me buy this product," and they can do it because they're a verified agent backed by a human. Things that don't fit the criteria.

This is a long list, but I just wanted to rule out many different careless cases that we see and often encounter. I won't dwell on this too much because I really want you to create whatever you want. And if you have any doubts about whether this is something completely normal that everyone builds, you can ask about it on our Discord channel. Static demos without any end-to-end integration with Agent Kit integration with Agent Kit are not what we need. Simple agent reputation, we've already seen many uses for this.

Um, commercial demos that only apply a discount to the agent, like we did here. This is something you can miss. We already did it. We need new and interesting uses for Agent Kit. Certain sample patterns for everything, really, not just Agent Kit.

You can use selfie verification to research funnels. So if your user is doing something, but you want only real people to have access to certain information, you can use selfie verification as a low- trust method to limit access to just them. You can create flows based on risks, for example: if you want users to access your platform, but you don't want to give them full access, and you don't want to do a complicated registration because it's boring for them and you might lose them you might lose them . You can, for example . You can, for example , ask users to pass a selfie verification to unlock a few features—low-risk ones, of course —and if they pass full verification, you can unlock riskier types of features or actions.

Ahem. And yes, free trials for selfie verification—that's also a very low threshold for entry, which has proven to be interesting and useful for many companies or startups that are currently using selfie using selfie verification. Because this again gives you an understanding: okay, there's a person behind this. I do n't know who it is. I don't need to know who it is.

So I can provide them with anything you want to provide without having to get their personal data. Regarding the requirements for the selfie- the requirements for the selfie- verification track: we want you to use it in a meaningful way, detailing developer and user feedback, preferably in a document. We need everything to work for you from start to finish. New verticals beyond the obvious examples I just mentioned. And you can use both V3 and V4 evidence.

V4 evidence will be available next week. For Agent Kit—the same is true for meaningful use, confirming that there is a human behind the agent before granting access to what you are creating. Again, a work- through flow. And of course, as I mentioned earlier, new verticals beyond the obvious examples. The evaluation of the results will be divided into these criteria.

Overall, is it strategically appropriate and beneficial for trust? Is this integration meaningful for the world or for our products? 25% our products? 25% will be allocated for how critical your reviews are. So don't be too polite.

We really want to test this out and are eager to hear what's bad, what's good, and what you would change. 20% for product quality, 15% for technical integration. We will check the correct integration with ID kit or agent kit, and 10% for deployment or feedback path. Could this become something more than just a hackathon project? Can you continue and are you going to develop this further?

This is, of course, for selfie check. That's all. This is a QR code for documentation. And for support, please support, please contact us in the Discord channel. I will stop the presentation and read the questions.

That's it. Oho. Long question. I tried agent signless. Good.

OK. Alexander. I think this is a long question. It would be ideal to discuss this on Discord where I can help you with the setup, as this will take some time. Okay, regarding the sandbox question: you need to access it first.

So, can selfie pay...can a selfie compare to a selfie already taken? So, do you want to have a selfie you've already taken and compare it with another selfie in real time using selfie check? Actually, this is what selfie check does. This is not entirely true, it uses an image stored on the device and compares it to the selfie we are taking at that moment.

Alexander, no, s cannot support an agent. Registration via agent kit is for orb verify only. That's why we have a sandbox. Uh, oh, I think I covered the selfie check ideas. But if you want more ideas, I'll be happy to send them to you or we can discuss it on Discord.

Sorry. What is the difference between selfie check legacy and proof of human? Are both suitable for selfies? Oh no, no, no. So, selfie check legacy.

Well, it's a matter of name. I think we need to change that. But selfie check legacy is selfie check, and proof of human is orb verified. Good. So, this is for the same SDK, i.

e. ID Kit. Two different credentials. We want you to build on the foundation of selfie verification. If you want to build on selfie verification and human verification, that's human verification, that's great, but you need to enable selfie verification.

Good. Can we see the QR code better? Uh -huh. Oh, yes, the QR code redirects you here. So you can go to the documentation from here, docs.

world.org. Uh-huh. And how we send our email for mailing and sandboxing. If you create an account, you will need to create it in our developer portal.

When you create an account there, you will see a button called " will see a button called " Access the developer portal" or "to the sandbox". You will request access to this . And on the World Prices page, let me find it quickly. You will be able to see links to forms. I also sent this form to the Discord channel, I think, an hour ago think, an hour ago .

But let me give you this. Here, hold on, guys. Is selfie verification supported indefinitely? No, maximum 90 days. 90 days.

And there you have World Tracks, and at the bottom in the resources section you will see access to the sandbox, there is a form there that you can use. And so, if you have no other questions, I think we can wrap up. Alexander, it would be great if you wrote to me on Discord. My nickname is like this. Here , please.

You can contact me on Discord. My nickname is Mr. Saon, because my last name is Saon, and it was a very obvious decision for me at 18. So yeah , that's all from my side. Thanks everyone, guys.