AI Safety
How we keep AI features safe and under human control.
AI is core to Blanca's Builder. This page describes the safeguards we apply when our product reasons, generates code or guides users.
Last updated: 2026-06-28
This page is maintained by Blanca's IT Professional SL to answer common security, privacy and trust questions about Blanca's Builder.
This page is app-owned editable content. It is not an independent audit or certification.
Security is a shared responsibility between Blanca's IT Professional SL, the platforms we run on (Cloudflare, Supabase) and you as the account owner.
Human in the loop
AI outputs are suggestions, never automatic actions. You review and accept changes before they are written to your project or deployed.
Scoped context
AI requests receive only the project context the user has chosen to expose: their prompts, the relevant memory entries and the active project brief. Other projects and other users' data are never included.
No training on your data
Where the provider supports it, we disable training on customer data. BYOK calls are governed by the contract you have with your chosen provider.
Prompt safety
Our system prompts forbid generating illegal content, doxxing, targeted harassment, malware and disinformation. The Pedagogical Mentor refuses requests that violate these rules and explains why.
Transparency
Generated content is labelled in the UI. You can inspect the prompts that produced any output, export them and delete them.
Canonical: https://blancasbuilder.com/trust/ai-safety · Blanca's Builder