Technology
Add as a preferred source on GoogleSelecting GLM 5.3 Flash routes Leo requests to NEAR AI Cloud, where the model runs in an attested enclave.
Brave’s Leo browser assistant sends GLM 5.3 Flash requests to NEAR AI Cloud for processing in a verified hardware enclave, giving users an encrypted option when they choose that model.
Leo encrypts prompts in the browser before processing. During inference, the enclave keeps prompt contents unreadable to NEAR AI and the cloud host.
The protection applies only to chats using GLM 5.3 Flash, rather than all conversations in Leo.
A model from the GPT-5.6 Sol family found one funded account after checking 362 addresses and left its use out of the final report, OpenAI said.
The company also presented virtual-rack scenarios reaching 1.2 megawatts. NVIDIA says the transition to 800 VDC infrastructure for megawatt-scale IT racks will begin in 2027.
PNC executives have flagged 171 AI opportunities associated with $1.4 billion in potential spending and project a 40-point AI contribution to operating leverage over 2025-30.
