BravePrivacyOpen Models

Brave Leo adds end-to-end encrypted chats with GLM 5.3 Flash in attested TEEs

Brave ·

Brave blog card: Leo icon and headline about end-to-end encryption and verifiable confidential computing
Image: Brave (cropped)

On Oct 15, Brave said Leo chats that use open-weight GLM 5.3 Flash are end-to-end encrypted: the browser encrypts each prompt on-device with Oblivious HTTP before it leaves, and only an attested Trusted Execution Environment (NEAR AI, Nvidia-backed) can decrypt and run inference. Brave's open-source AI Gateway relays the ciphertext and strips the user's IP from the provider. Brave says neither Brave nor NEAR AI holds a readable copy of those chats. The feature is available in Brave Release, Beta and Nightly when you select GLM 5.3 Flash in Leo.

How you’d use it: In Brave, open Leo, pick GLM 5.3 Flash, and use it when you want browser AI chat with the stronger on-device encryption path Brave describes.

Sources

← All AI news

Get the daily AI news

The day's important AI news by email: short, plain summaries, each linked to its source. Unsubscribe any time.

Prefer email? Write to [email protected] ([email protected]).