On October 15, 2026, Brave said Leo, the AI assistant built into its browser, now offers end-to-end encrypted chats. Brave co-founder and CEO Brendan Eich was one of four authors of the announcement. The feature works with one model for now: GLM 5.3 Flash, an open model from the company Z.ai.

What Brave says happens to your words

Brave says your browser locks each message on your device before it is sent. Brave says only a sealed, checked piece of hardware can unlock it. That hardware is run by a company called NEAR AI and uses Nvidia chips. Brave says neither it nor NEAR AI can read your chats, and that Z.ai, which made the model, never sees them.

Brave says its own servers pass the locked message along and remove your IP address on the way, so the AI provider does not learn who is asking. Brave says your chat history stays on your device.

Why "verify" matters here

Brave's pitch is that you should not have to take a company's word for privacy. It says the hardware must prove what code and which model it is running before your browser will send anything. Brave says the code for each step is public so outsiders can check it. For now, Brave says, your browser relies on Brave's server to do that check. It says moving the check into the browser itself is a next step, not something shipped today.

What you can try

Brave says this is available now in all versions of the browser: Release, Beta and Nightly. Open Leo, choose GLM 5.3 Flash, and chat as usual. Brave calls the GLM 5.3 Flash option a beta, and says more Leo features and models will come to private mode later.

Source: Brave