AI File Renaming Privacy and Security

Zush is designed to rename files by analyzing the minimum useful content for the selected workflow. What is processed depends on the AI mode and file type.

Zush Cloud AI and BYOK processing

In Zush Cloud AI and BYOK modes, Zush may process an analysis payload such as:

compressed image or design previews
sampled video frames
subtitle or transcript context
extracted document or PDF text
temporary iWork preview output when needed for analysis
audio metadata, transcript context, or bounded audio payload when transcription is required
MIME type, extension, language preference, regenerate settings, and custom prompt text

Original full-resolution files and local folder paths are not sent as separate fields in normal cloud operation, except where a bounded audio payload or temporary document preview is required for analysis.

Cloud providers

Managed processing can use Google Gemini, OpenRouter and its upstream providers (including DeepSeek), Alibaba Cloud Model Studio (Qwen), and Groq for speech transcription. The route can differ between file analysis, fallback analysis, and folder planning.

Qwen requests may use a Global service deployment scope where encrypted transient inference data can be processed outside the selected US region, including within or outside China; static request data remains in the selected US region.

Local AI processing

Ollama and LM Studio use models managed by those local applications. In both modes, supported file analysis content is processed on your Mac or Windows PC instead of by Zush Cloud AI or a third-party cloud AI provider. Zush may still contact backend services for licensing, updates, support, or operational checks.

BYOK keys

BYOK keys are stored locally in secure platform storage. In BYOK mode, the key is used to validate the provider and process selected AI requests. Zush does not permanently store BYOK API keys.

Training

Zush does not use your file content or custom prompt text to train Zush-owned models.

Regulated documents: medical, legal, financial

If you handle documents under confidentiality or regulatory obligations — patient records, client files, financial statements — choose the AI mode that matches your posture:

Ollama and LM Studio process supported files on your Mac or Windows PC with local models. Nothing about the file content is sent to a cloud AI provider while either mode is selected.
BYOK sends analysis through your own Google Gemini, OpenRouter, OpenAI, or Anthropic account. When OpenRouter is selected, it routes the request to the upstream provider serving the selected model, so review both providers’ terms for that workflow.
Cloud mode uses the managed providers described above. Choose a local mode instead when file content must stay in-house.

File operations happen on your disk. A rename changes the filename; choosing a destination or enabling folder sorting can also move the file. Zush does not store your original files as a hosted library. Cloud and BYOK processing send the analysis content described above; local AI modes keep supported analysis content on your computer.

Two practical caveats. First, an AI mode does not replace your organization’s security, access, retention, or naming policies. Second, filenames can surface in sync logs, attachment lines, and search indexes, so keep details that do not need to appear in a folder listing inside the document. The medical records workflow shows how to combine this with local processing or organization-controlled BYOK.

Read the full policy

This page is a practical product summary. For the legal version, read the Privacy Policy.