AI Data Processing & Training

Last updated 20 October 2026 · VantyrixTek

What is sent to AI providers

To fulfil your request, SHOUD sends relevant prompts, project instructions, preferences, prior conversation context and supported source-file contents to Gemini through Google Cloud. A separate bounded review request may process the draft and task. Research can send derived search queries and context to Google Search. Do not upload information you lack authority to disclose.

Training and improvement

The current SHOUD implementation does not train or fine-tune AI models on customer prompts or files. Google’s published managed-model terms restrict training or fine-tuning on customer data without prior permission or instruction. This is a provider-terms disclosure, not a claim that SHOUD has negotiated a special zero-retention agreement. Read Google’s current data-governance guidance.

Retention is different from training

Provider safety logging, caching, grounding and operational processing may retain information under applicable settings and provider terms. SHOUD does not promise zero retention or region-exclusive AI processing. Our task prompts, responses, source files and usage history are separately stored for the workspace as described in retention. Training restrictions do not mean a request is never stored.

Sensitive and business data

Avoid payment credentials, secrets, children’s records and regulated sensitive information unless you have a lawful basis and confirmed safeguards. Business customers needing documented instructions, transfer safeguards or a processor agreement should arrange a DPA before submission. Google Analytics is not configured to receive your prompts or uploaded file contents.

Read these notices together with the Terms and Privacy Policy. Mandatory legal rights remain applicable. Questions: bolt@omnibiofex.cloud.