How BoodleBox handles your prompts
Goal
Understand what happens to a prompt between you and the AI model, and what leaves BoodleBox along the way.
Your identity is replaced first
Pseudonymization is the first step when a request enters BoodleBox. Your name, email and institution are replaced before anything else happens, so no AI provider or tool receives them.
How requests reach AI models
Requests go from BoodleBox's servers to the AI provider, never directly from your browser.
Most go through Microsoft Azure AI Foundry. Some models use the provider's own API (for example Google Gemini), and AWS Bedrock in US regions is the backup route.
Providers include Anthropic, OpenAI, Google and others.
Tools: web search, image generation and model tool calls
Tools run after pseudonymization.
Web search queries and image prompts are prepared inside BoodleBox and sent to the search or image provider without your identity.
When a model asks to use a tool, the request comes back to BoodleBox and runs there. Only the tool's output goes back to the AI provider.
BoodleBox's built-in bots
Our bots run on the same AI models through the same path, so the same pseudonymization applies. They aren't autonomous agents: a person starts and controls each step.
What pseudonymization doesn't cover
We say pseudonymization, not anonymization, on purpose. It covers who is asking, not what you type.
BoodleBox keeps the link between a request and you, so it can return the response and save it to your history.
What you type, and the files you attach, are sent to the model as written. Personal information you include in a prompt or file isn't removed.
Institutions: we recommend covering this in your acceptable use guidance. See Data Governance & Classification for a model of what data belongs in AI tools.
What AI providers can do with it
Providers are contractually prohibited from training on customer data. We can confirm this in writing.
Our primary provider, Azure AI Foundry, keeps data for under 30 days. Other providers may keep prompts for up to 30 days under their own terms (see Understanding data retention policies).
Your data at rest
We collect minimal personal data: your name, email, institution or role, and the content you create.
Storage is US-only, and retrieval is scoped to your organization, so one institution's data isn't searched for another's.
Documentation
Our SOC 2 Type II report, subprocessor list and supporting documents are on the Trust Center: https://trust.boodlebox.ai/ (some items require an NDA). For a diagram of how prompt data flows, email compliance@boodle.ai.
Questions?
Security and privacy: compliance@boodle.ai. General help: reach out through our contact form.