The core problem OpenGradient solves
Most AI assistants (ChatGPT, Claude, Gemini, etc.) tie your questions to your identity, store them on servers, and one data breach away from being public.
OpenGradient chat's privacy works in 3 layers.
Layer-1
Encrypted on your device, so the keys live only on your device.
Layer-2 oblivious HTTP Relay:
The relay sees your IP but only encrypted data. So neither party alone can connect both pieces. That's the key point — no single party sees the full picture.
Layer-3 Trusted Execution Environment:
Your prompt is only decrypted inside a sealed, attested enclave. So I can say that the operator cannot read or log your prompts and that's what makes it powerful.
To conclude that the privacy is in the architecture of @OpenGradient , not just a policy.