# Does ZeroTwo use my chats to train AI models?

> What the ZeroTwo Privacy Policy and in-app disclosures say about AI providers using submitted data, and what you can do about it.

Source: https://help.zerotwo.ai/en/articles/5722486-does-zerotwo-use-my-chats-to-train-ai-models
Section: Privacy and policies > Policy FAQ
Last updated: 2026-10-05

The **Data Controls** page in ZeroTwo's settings has export, archive, delete and shared-link controls, but no switch for model training. What happens to a message depends on the AI provider behind the model you chose: the [ZeroTwo Privacy Policy](https://zerotwo.ai/privacy) says most providers only process content to return a response, and that some may keep or use it under their own terms.

## What does the Privacy Policy say?

- **Your content goes to the provider you select.** To provide AI features, ZeroTwo may send that provider your prompts, the chat history needed for the current task, files or media you attach, retrieved snippets from your workspace, project instructions, and profile or memory context you have enabled.
- **Providers differ.** Most process submitted content only to generate a response. Some may retain or use it under their own terms or privacy notices, so the policy tells you to review the provider-specific disclosures shown in ZeroTwo and the provider's own policies before using a model for sensitive work.
- **Connected services are excluded.** ZeroTwo states that it does not use data from Google Workspace APIs or from connected third-party integrations to train generalized AI or machine learning models, and that this data is used only for the features and automations you request. See [How ZeroTwo handles your Google account data](https://help.zerotwo.ai/en/articles/10408842-how-zerotwo-handles-your-google-account-data).

For your chat content itself, the policy describes sending it to the provider you select and does not describe a training opt-out. The providers it names are Anthropic, OpenAI, Google, xAI, Cohere, Mistral, Perplexity, Groq, fal, Stability AI, DeepSeek, Qwen and Kimi, and it says others may be added over time.

## Which models does ZeroTwo describe as training on prompts?

ZeroTwo's model catalog includes two Meta models named "Muse Spark 1.3 (c)" and "Muse Spark 1.2 (c)". They are the Contributor-tier versions of Muse Spark 1.3 and 1.2: the same checkpoint at a discount, in exchange for permission for Meta to train future models on your prompts and completions. The catalog tells you to use the Standard tier for anything confidential.

## What do the iOS and Android apps tell me?

Both mobile apps have an AI data sharing disclosure. It lists what may be sent, including prompts, assistant replies, attachments and images, retrieved file snippets, project instructions and the profile or memory context you have enabled. It adds that most providers only process the data to return a response, and that "Kimi, DeepSeek, and Qwen may use submitted data for training."

- **iOS:** a "Before you start chatting" screen shows the disclosure, and you tap **Allow and continue** to let ZeroTwo share chat data with third-party AI providers. You can read it again under **Settings** > **AI data sharing consent**.
- **Android:** open **Settings** > **AI data sharing consent**. The screen is titled "AI Consent".

## Is Memory the same as training?

Memory is a different feature. ZeroTwo describes it as what it carries from one chat to the next, stored with your ZeroTwo account and used in every chat while Memory is on. While it is on, saved memories can be part of the context sent to the provider with your message. For the Memory controls, see [How does ZeroTwo handle my data?](https://help.zerotwo.ai/en/articles/7039943-how-does-zerotwo-handle-my-data)

## What can I do to limit how my content is used?

- **Choose the model deliberately.** The provider behind each model sets the rules, so use a model whose provider you have checked for sensitive work. The mobile disclosure names Kimi, DeepSeek and Qwen as providers that may use submitted data for training, and the catalog describes the Muse Spark Contributor tier the same way.
- **Leave sensitive content out.** The policy puts this in your hands: if you do not want certain information processed through ZeroTwo or third-party AI providers, do not submit it.
- **Turn Memory off** in **Settings** > **Personalization** > **Memory** on the web or desktop app. When it is off, nothing is used or learned.
- **Delete what you no longer need** from Data Controls. See [Manage your chats and account data in Data Controls](https://help.zerotwo.ai/en/articles/7730893-manage-your-chats-and-account-data-in-data-controls). Deleting removes content from your ZeroTwo account; the policy notes that some providers may retain submitted content under their own terms.

If you need a firmer answer for a specific situation, email [reed@zerotwo.ai](mailto:reed@zerotwo.ai), the contact address in the policy.