Untold and Venice, side by side

Both are private AI. Both name the same four privacy modes and attest against the same enclave provider. What follows is what is the same, what is different, and where each line was read — every claim about Venice links to their own page and the date we read it, and every claim about us links to the page here that proves it.

What is the same

  • Four privacy modes with the same names. Anonymized, Private, TEE, and End-to-End Encrypted — theirs described as identity stripped, contractual zero retention, hardware enclaves, and client-side encryption. venice.ai/privacy, read 2026-09-11 Ours are the same four, and the badge on every answer names which one it got — see the models.
  • The same enclave provider. Both attest against Phala for the TEE tier — theirs described as “hardware-secured enclaves” venice.ai/privacy, read 2026-09-11, ours with the enclave named on every sealed answer’s receipt (see the models). Two of the models are the same models; we measured this against both live model lists rather than reading it in a blog post.
  • Conversation history on your device. Theirs: “Your conversation history stays on your device. Venice never stores it on its servers.” venice.ai/privacy, read 2026-09-11 Ours: a switch in the composer’s privacy choices does the same, and says what it costs.
  • Memory kept in the browser. Their Memoria stores memories and documents in IndexedDB on your device. Memoria technical overview, read 2026-09-11 Ours does too. One difference sits in the next section.
  • The same prices. Free, $18, $68, $200 a month. venice.ai/pricing, read 2026-09-11 Ours are the same four numbers — see plans. What differs is what the paid tier buys: on both, it is topic freedom.

What is different

  • A receipt you can check yourself, after the fact. Every sealed answer here — and an agent answer whose every model call the enclave proved, one call or a run of several — carries a signed receipt naming the enclave that produced it, and the drawer under the answer lets you verify it offline, link by link; the check makes no network call and does not ask us. If the enclave cannot prove one of an agent run’s calls, the answer is not served at all — the run tries another model, or refuses and says which call failed (see the limits). Venice publishes that its TEE tier is “verified by remote attestation” venice.ai/privacy, read 2026-09-11; we did not find a per-answer receipt handed to the person who asked. The one-command verifier package is not published yet, and our API page says so.
  • A refusal that says what happened. The most consistent complaint in reviews of Venice is that a blocked request comes back as “model not available” rather than a refusal. Reddit review roundup, read 2026-09-11 That is a report from users, not a measurement of ours. What we can say about our side: an answer that never arrived says so, a stop says it was a stop, and a request the service could not complete names the reason.
  • A page of what we do not promise. /limits is generated from the running system and linked from the footer of every page. We did not find an equivalent on Venice’s site; if one exists, tell us and it goes here.
  • Where the document text goes to be embedded. Venice’s own technical overview: “Venice servers only generate embeddings transiently” — the document leaves your device to be turned into vectors, then the vectors come back. Memoria technical overview, read 2026-09-11 We are building embedding that runs in the browser, so the text never leaves at all. It is not shipped. It is on this page because it is the difference we are working toward, and because saying so is cheaper than being caught implying it.

Where Venice is ahead

  • Image, video and audio generation are live there. venice.ai, read 2026-09-11 On this deployment, image and video are built and waiting on a provider; the studio says so on the page rather than showing you a button that does nothing.
  • Mobile apps. They have them. App Store listing, read 2026-09-11 We are a web app that works on a phone.
  • Model breadth. Dozens of image models across their two lanes. Our catalogue is smaller and every entry on it has been probed for a receipt — see the models.
  • A public feature-request board. Theirs is public. Ours is this page’s footer: tell us.

How this page was made

Every quotation from Venice was read from the linked page on 2026-09-11. If any of it has changed since, the link is there to check, and the correction is one message away. The claims about us link to the pages that carry them, and the two things we have not shipped are labelled as not shipped.