AI Policy

Blanca's Builder routes AI requests to one of: the Lovable AI Gateway (managed mode), the provider whose API key you supplied (BYOK mode — OpenAI, Anthropic, Google, Mistral), or a self-hosted endpoint you configure. You can see and change the active provider at any time in AI Settings.

We send only the data required for the request: your prompt, the relevant project context from your Brain & Memory Bank, and the system prompt that defines the AI's role. We do not send other users' data, billing data or unrelated projects.

We do not train AI models on your data and we configure each provider so that they do not train on it either, where the provider offers that option. In BYOK mode, the terms of your chosen provider apply directly to you.

AI output is suggestive, not authoritative. You remain responsible for reviewing generated code, copy and decisions before publishing. The Pedagogical AI Mentor and AI Coach are designed to teach, not to replace your judgement.

You may not use our AI features to generate content that is illegal, that infringes intellectual property, that targets or impersonates a real person without consent, that spreads disinformation, or that is intended to manipulate, deceive or harm others.

AI-generated content is clearly labelled inside the product. You can export your prompts and outputs at any time and you can delete them from your project history.