Why we chose ZEC as the payment layer for Silo, our private AI platform

Zcash Apps
1

We’ve been following the Zcash community for a while. The commitment to building real, verifiable privacy rather than just claiming it is something we deeply respect and try to hold ourselves to as well. When it came time to solve the payment layer problem on Silo, ZEC was the only project we trusted to do it right.

Here’s the problem we kept running into: most “private” AI tools put real effort into the query layer (secure enclaves, anonymization proxies, PII stripping) and then let you pay with a credit card. Stripe knows who you are. Your bank knows. There’s a timestamped receipt in a log that says you used this service, when, and how much you spent.

For a lot of use cases that’s fine. But private AI exists specifically for the cases where it’s not fine. Medical, legal, journalism. Situations where even the fact that you’re researching something is sensitive. A payment trail breaks the whole thing. ZEC’s shielded transactions don’t leave a readable trail. That’s why we’re here.

For anyone curious about the full stack we’ve built on Silo (siloprivacy.com):

  1. Open-source models (Kimi K2.5, etc.) run inside TEEs. No API calls leave the enclave, attestation is verifiable, we can’t see your data either.

  2. For closed-source models like GPT that require API calls, we pool queries through proxy servers and run them through an anonymization model we built (Anonymizer-4B on HuggingFace) that strips PII before anything gets submitted. Not perfect but meaningfully better than nothing.

  3. Private Deep Research: the full reasoning chain runs inside a series of secure enclaves with only encrypted messages passed between them. If you use deep research for anything sensitive, this matters a lot.

  4. For people who want no cloud dependency at all, we have a hardware box you can run at home. 288 GB GPU RAM, 512 GB system RAM. Overkill for most people but some users really want it.

Payment options are Stripe/Apple Pay if you don’t care about privacy on that layer, ZEC if you do, or our own FAI token for discounted access.

We’re glad to be supporting ZEC and hope this is the start of a longer relationship with this community. Happy to answer any questions about the architecture or how the ZEC integration works.

– The Freysa team

7 Likes