AI and your data

1. Our commitments

When you ask Ambicast’s AI to redesign a document, your content is used to do that one job for you and for nothing else.

• Never used to train AI. Not by Ambicast, and our AI providers are barred by their own terms from training on your content.
• Never sold or shared. Your content is processed only to do the job you asked for. It is never sold, and never shared with anyone for their own purposes.
• Not kept by us. We do not store your questions, your documents or the text read from them. Only the slides you create are saved, in your workspace.
• Sent only when you ask. Adding a file only reads its words. Nothing goes to the AI until you press the button to start.

2. What is sent to the AI

The design wizard

The words you type or paste, or the words we read from a file you upload (for example a PDF or a text file), plus the style choices you make. We read a file on our server only to pull the words out of it. The file itself, and any pictures in it, are not sent to the AI.

The AI Editor

Your question, and the details of your workspace that the assistant looks up to answer it, such as screen names, playlist names, schedules and campaigns. It can look things up. It cannot change anything.

The Studio assistant and text tools

The instruction you type and the text you select. The Studio assistant also sees the names, positions and short excerpts of the text on the layers of the slide you are working on, and the names of your screens.

Your payment details and passwords are never sent, and neither are the pictures and videos in your media library. The AI is only given what your request needs.

3. What happens, step by step

1. 1You add your document. For a file, our server reads the words out of it in memory so you can see the text. The AI is not involved yet, and the file is not saved by the AI tools.
2. 2You press the button to start, such as Design it. Only now is the text sent, over an encrypted connection, to the AI model.
3. 3The model works out the structure and wording and sends the result back. It processes your content only to answer this request, on our behalf.
4. 4We turn the result into slides in your workspace. Nothing goes live on a screen until you publish it.
5. 5Our usage records keep how much AI was used, as a number, and never what was in your request. Our logs record errors, not your documents or questions, and our error monitoring does not record your screen or session.
6. 6The AI Editor’s chat is not saved. It lives in your browser tab and is gone when you close it.

4. Who runs the AI

Two providers run the models. Both are listed on our Sub-processors page, and both are bound by terms that keep your content out of model training.

• Anthropic runs the Claude models behind the AI Editor and the design wizard, which is where your documents go. Under Anthropic’s commercial terms it may not train its models on the content we send, and that content is treated as confidential information. Anthropic is based in the United States, so your text may be processed outside the UK.
• Amazon Web Services (Amazon Bedrock) runs the quick text tools, such as rewriting, and the Studio assistant, in London, on Amazon’s own models. It is also our fallback if Anthropic refuses a request. AWS states that it does not use inputs or outputs to train any model and does not share them with model providers.

Both act as our processors: they handle your content to provide the service to us, and for the safety checks described below, not for their own purposes. How transfers outside the UK are protected is set out in our Privacy Policy and Data Processing Addendum.

5. How long is anything kept?

Ambicast does not keep a copy of your AI requests or of the document text. What the AI produces becomes slides in your workspace, like any playlist, and stays until you delete it.

Our AI providers may keep a request briefly to run safety checks, then delete it. Anthropic currently deletes inputs and outputs within 30 days. It may keep a request longer only where the law requires it, or where its automated safety systems flag a possible breach of its usage policy. Amazon Bedrock may retain requests for abuse detection on some models; AWS states that anything it retains stays within AWS and is not shared with the model provider.

These are the providers’ own published policies and they can change, which is why we name them here and keep the list on our Sub-processors page up to date.

6. If your document is sensitive

• Take out what the slide does not need. Names, account numbers and private figures can be swapped for [placeholders] before you press the button, then filled in inside the editor.
• Do not include passwords, payment card numbers or medical records.
• You stay in control of what goes live. The AI drafts; nothing appears on a screen until you publish it.
• The AI features are optional. They only run when you press a button or send a question, so if you never use them, nothing is sent to an AI. You can also connect your own AI assistant from Settings, Developers, and use it under your own agreement.

7. Questions

For a security review, a copy of our Data Processing Addendum, or any question about this page, email privacy@ambicast.tv. Your rights over personal data in our Privacy Policy apply to anything the AI features handle.

This page explains how our AI features work today. The binding terms are in our Terms of Service, Privacy Policy and Data Processing Addendum.