90% of AI apps ask for your trust

Chances are, if you’ve used an AI app you’ve had to place trust in the operator.

Trusting the operator is the usual and expected case in portfolio management, lending, medical information services and content moderation.

There’s no independent way to verify whether the output that you received was logged or tampered with.

Where privacy means obfuscation

There are some apps which go further by offering encryption. But encryption only hides what’s sent.

Operators can still see what’s in the prompt, make a copy of the encrypted message and decrypt it for their use at any time.

In a world where prompt injection is an art, this is a big problem.

OpenGradient wants to replace trust with proof.

Privacy by cryptography

- The messages you write are encrypted on your device.

- No one sees anything in the message before it’s encrypted.

- Your identity gets stripped off before anything gets sent to the model.

- Your privacy is enforced by cryptography and hardware, not promises.

You can be certain that no one sees the message at any point along the route to the model.

Verifying inference is hard, And most apps let inference happen in one black box.

And you have no way to independently verify that inference was done correctly, and there’s no way to verify that your prompt was the one used.

A few experimental projects come close by making computation verifiable.

But OPG takes this further by running the verified inference network on a permissionless network of specialized nodes.

through it's TEE-based secure and verifiable LLM inference

OpenGradient delivers verifiable LLM inference for running LLMs on the network. All LLM prompt requests are routed through the Trusted Execution Environments (TEEs) which provide hardware attestation verification meaning node operators can’t see, log or manipulate requests.

You can be certain every inference is executed correctly, your prompt is the one being used and no one sees what you wrote.

Why it matters

This clearly matters for many sensitive tasks, like

- Medical reasoning

- Financial analysis

- even private conversations

All of these would be difficult to use with today’s AI apps.

And you can do it all from one app.

OPG also gives you access to all of the major frontier models in one app.

Claude’s latest Fable 5 model now currently integrated too and The Nous Hermes model is available in Private Chat.

You get private, TEE-verified inference through a unified API.

Users who have been actively using the chat platform with purchased credit will be eligible for the S2 $OPG airdrop too.

If you want an AI you can trust with anything, check out OpenGradient Chat, new users are get access to 1,000 free credits upon signing up: chat.opengradient.ai