AI notice

Last updated: 9 November 2026

You are talking to AI

Every answer in cypher is written by an AI model, not by a person.

Which models

For now, every answer comes from Anthropic’s Claude models. Each answer shows the model that wrote it, and the app’s model menu lists the ones you can pick.

How Auto picks a model

When you choose Auto, cypher picks the model for your chat. For now, that is Claude Sonnet, with Claude Haiku as its backup; next, Auto will pick by the kind of task, from a routing table we maintain. A chat then keeps its model. cypher moves it to another only in one of these cases:

• The chat no longer fits the model.
• A message needs something the model can’t do, for example read an image.
• The model is retired.

With Auto, cypher tries Claude Sonnet first, and Claude Haiku if Sonnet fails before an answer starts. The model that gives a new chat its first answer becomes the chat’s model. After that, if the chat’s model fails before an answer starts, the other of the two answers that one message, and the chat keeps its model. You can always pick a model yourself.

Long chats

When a chat gets long, an AI model summarises its earlier messages, and later answers get the summary instead of every word. You still see all your messages.

What AI gets wrong

• Answers can be wrong, out of date or made up, and still sound sure.
• Answers can carry the biases of the data the models learned from.
• Models don’t know about events after their training unless you tell them.
• Don’t rely on an answer alone for health, legal, financial or safety decisions.

AI-generated content

Text that cypher writes is AI-generated. If you publish it where people could think a person wrote it, and that matters to them, say that it was made with AI.

Your chats and the providers

The privacy policy says which providers receive your messages, how long they keep them, and what their terms allow them to do with them.

Questions and reports

To report a harmful or wrong answer, write to hello@gocypher.ai.