{"id":"Shxc59Rkg68","title":"Brave Leo’s New AI Model Can Prove It’s Private","description":"Brave Leo can now run GLM 5.3 Flash on NEAR AI Cloud, inside hardware enclaves that keep your prompts sealed while the model runs. Every response comes back with an attestation you can check. In this video I explain how it works, then show you how to turn it on in about a minute.\n\nMost AI assistants protect your data with a privacy policy. Your prompt is still readable at runtime by logging systems and the people running the servers. With NEAR AI's trusted execution environments, the cloud host, the GPU operator and NEAR AI itself are locked out at the hardware level. Brave adds its own protections around the model too: no IP collection, no prompts kept on its servers, no training on your chats, and no account required.\n\nThe blood work question in the demo uses made-up numbers.\n\nChapters\n0:00 Private AI inside your browser\n0:14 Leo's new model on NEAR AI Cloud\n0:54 Privacy policies vs. hardware proof\n1:10 Leo's privacy harness\n1:55 How a request stays sealed\n2:16 Download Brave\n2:36 Open Leo AI\n2:50 Switch to GLM 5.3 Flash\n3:05 Ask a private question\n3:30 Check the NEAR verified badge\n3:50 Get started\n\nLinks\nDownload Brave: https://brave.com/download/\nNEAR AI: https://near.ai\nBrave's privacy policy for Leo: https://brave.com/privacy/browser/#br...\nMore private AI videos:    • Building with AI  \n\n#NEARAI #BraveLeo","channel_name":"NEAR DevHub","published_at":"2026-10-15T00:00:00.000Z","thumbnail_url":"https://img.youtube.com/vi/Shxc59Rkg68/hqdefault.jpg","transcript_text":"When you use an AI system inside your browser, you actually know what's happening with that information. Well, Brave is a privacy first browser and they are focused on protecting user data. And now their Leo AI has gotten even better because it uses Near AI which has privacy attested inference. I'm excited for this. Let me show you why. Reso now gets a model that can actually prove that it's private. You can use GLM 5.3 flash on near aa cloud which runs inside hardware enclaves which means that every response that's done by the inference comes back with a stationation that you can actually check now allows Leo to get answers from an actual sealed enclave. What that means is that no one can actually see what's actually being processed. So near AI cannot see it. The actual cloud providers or any other data center can't see it either. the process is actually happening inside the trusted execution environment. The reason why this is special and unique is that most assistants will actually just give you a privacy policy. Privacy policy just tells you how the company may or may not handle your information, but your prompt is still readable at runtime by any logging systems and the people who are running those different servers. But LEO on a neari cloud now has hardware isolation. So you get a signed addestation on every response. The other benefit here is that Leo is actually putting its own harness around the model, which is a privacy firstbased harness. Ray doesn't collect your IP address or identify you any other way. There's no prompts or responses kept on its server. This can train models on your conversation. Your chat says on your device or if you sync under your own key and it's free. You don't need an actual account. Combine that with a trusted execution environment on your AI cloud, you have a really strong privacy first chatbot inside your browser. The setup here is actually really simple. You can see that Brave Leo runs inside your browser. It has the encrypted request. It goes to the trusted execution environment on your iCloud. It has confidential compute and the prompt is decrypted only inside there. Memory stays encrypted while the model runs so no one can actually see what was being asked. If any of this seems too complicated, no problem. Let me just show you how you can get started today and use private inference inside your browser. First, if you haven't already, go to brave.com and you can go and download the version of the browser for whatever system you're on, whether it's Linux, Mac, Windows, whatever, you can actually do that really easily from here, just like any other piece of software. I've opened up Brave on my computer and you can see over here there's these little stars and that is for the AI system or I can even hit the hamburger menu and also go to Leo AI. It will open up in the side and I can even open it up in full screen if I would like. To switch it to near AI is really really simple. I'm just going to go down here to the bottom. Switch from automatic to GLM5 flash. You can see the near logo here. verifiably private with near a this is more information what like I just mentioned here before I will select this and now I can trust I'm going to ask it a question and I know it's private so I'm going to put in a fake question this is not my actual health information but I'm going to ask a question about my blood work and see what it says and know that it's actually private all the way through all right so some answers coming back but the beauty of all of this is that I know that it's been actually protected and verified. So I go to the bottom here and I can see GLM 5.3 flash near verified and lets me know that it used that on that model and I have all this information now that is really helpful to me and I know it's private. As you can see it's really easy to get started. So download Brave and try out Leo AI with Near AI inference today. And for more information you can go to near.ai. Learn all about the different models and private inference provided by Near AI.","transcript_synced_at":"2026-11-03T03:46:19.612Z","segments":[{"start_seconds":0,"text":"When you use an AI system inside your browser, you actually know what's happening with that information. Well, Brave is a privacy first browser and they are focused on protecting user","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=0s"},{"start_seconds":8,"text":"data. And now their Leo AI has gotten even better because it uses Near AI which has privacy attested inference.","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=8s"},{"start_seconds":14,"text":"I'm excited for this. Let me show you why. Reso now gets a model that can actually prove that it's private. You can use GLM 5.3 flash on near aa cloud","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=14s"},{"start_seconds":23,"text":"which runs inside hardware enclaves which means that every response that's done by the inference comes back with a stationation that you can actually check","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=23s"},{"start_seconds":32,"text":"now allows Leo to get answers from an actual sealed enclave. What that means is that no one can actually see what's","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=32s"},{"start_seconds":39,"text":"actually being processed. So near AI cannot see it. The actual cloud providers or any other data center can't","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=39s"},{"start_seconds":46,"text":"see it either. the process is actually happening inside the trusted execution environment. The reason why this is special and unique is that most assistants will actually just give you a","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=46s"},{"start_seconds":54,"text":"privacy policy. Privacy policy just tells you how the company may or may not handle your information, but your prompt is still readable at runtime by any","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=54s"},{"start_seconds":63,"text":"logging systems and the people who are running those different servers. But LEO on a neari cloud now has hardware isolation. So you get a signed","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=63s"},{"start_seconds":70,"text":"addestation on every response. The other benefit here is that Leo is actually putting its own harness around the model, which is a privacy firstbased harness. Ray doesn't collect your IP address or identify you any other way.","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=70s"},{"start_seconds":81,"text":"There's no prompts or responses kept on its server. This can train models on your conversation. Your chat says on your device or if you sync under your","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=81s"},{"start_seconds":88,"text":"own key and it's free. You don't need an actual account. Combine that with a trusted execution environment on your AI","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=88s"},{"start_seconds":95,"text":"cloud, you have a really strong privacy first chatbot inside your browser.","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=95s"},{"start_seconds":115,"text":"The setup here is actually really simple. You can see that Brave Leo runs inside your browser. It has the encrypted request. It goes to the","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=115s"},{"start_seconds":123,"text":"trusted execution environment on your iCloud. It has confidential compute and the prompt is decrypted only inside there. Memory stays encrypted while the","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=123s"},{"start_seconds":131,"text":"model runs so no one can actually see what was being asked. If any of this seems too complicated, no problem. Let me just show you how you can get started today and use private inference inside","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=131s"},{"start_seconds":140,"text":"your browser. First, if you haven't already, go to brave.com and you can go and download the version of the browser for whatever system you're on, whether","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=140s"},{"start_seconds":148,"text":"it's Linux, Mac, Windows, whatever, you can actually do that really easily from here, just like any other piece of software. I've opened up Brave on my","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=148s"},{"start_seconds":156,"text":"computer and you can see over here there's these little stars and that is for the AI system or I can even hit the","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=156s"},{"start_seconds":162,"text":"hamburger menu and also go to Leo AI. It will open up in the side and I can even open it up in full screen if I would","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=162s"},{"start_seconds":170,"text":"like. To switch it to near AI is really really simple. I'm just going to go down here to the bottom. Switch from","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=170s"},{"start_seconds":176,"text":"automatic to GLM5 flash. You can see the near logo here. verifiably private with near a this is more information what","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=176s"},{"start_seconds":185,"text":"like I just mentioned here before I will select this and now I can trust I'm going to ask it a question and I know it's private so I'm going to put in a","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=185s"},{"start_seconds":193,"text":"fake question this is not my actual health information but I'm going to ask a question about my blood work and see what it says and know that it's actually","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=193s"},{"start_seconds":201,"text":"private all the way through all right so some answers coming back","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=201s"},{"start_seconds":210,"text":"but the beauty of all of this is that I know that it's been actually protected and verified. So I go to the bottom here and I can see GLM 5.3 flash near","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=210s"},{"start_seconds":219,"text":"verified and lets me know that it used that on that model and I have all this information now that is really helpful","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=219s"},{"start_seconds":226,"text":"to me and I know it's private. As you can see it's really easy to get started.","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=226s"},{"start_seconds":230,"text":"So download Brave and try out Leo AI with Near AI inference today. And for more information you can go to near.ai.","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=230s"},{"start_seconds":237,"text":"Learn all about the different models and private inference provided by Near AI.","url":"https://www.youtube.com/watch?v=Shxc59Rkg68&t=237s"}],"youtube_url":"https://youtube.com/watch?v=Shxc59Rkg68","url":"https://near-pedia.xyz/videos/Shxc59Rkg68"}