AI Disclosure

1. What is AI here

Kansumi uses third-party AI models to read floor plans and sketches, generate and edit interior images, and answer chat messages. Depending on configuration, requests are processed by fal, OpenRouter, Groq, OpenAI, Anthropic or a model self-hosted by the operator of this deployment, together with the model developers behind them; providers and models can change. Interior renders, including initial concepts, regenerations, style changes and edits, are AI-generated images. They are not photographs and not the work of an architect, engineer or designer. The assistant is automated. Data handling is described in the Privacy Policy.

2. What can be wrong, and what to trust

AI output can be inaccurate, inconsistent between runs, unsuitable, or impossible to build. What to trust:

• The 2D plan is the measurement. Renders take cues from it but are not dimensionally reliable.
• Extraction is AI-assisted. Walls, doors, windows and scale detected from a drawing can be wrong. Confirming one dimension sets the scale but does not verify the rest, so check and correct the scene before generating.
• Photo projects have no plan. Projects created from a room photo produce no measured plan.
• The assistant is not verified. Its answers about fit, clearance, products, prices or rules are probabilistic.

None of this is professional or safety advice; the Terms of Service set out the rules on construction and professional reliance.

3. How output is labeled

Stored renders and PNG exports are designed to carry signed C2PA Content Credentials that identify them as AI-generated; PDF exports keep the credentials inside the embedded page images. Free-tier renders and exports also carry a visible "Made with Kansumi" watermark; Pro renders normally do not. Credentials and watermarks can be lost through screenshots, messaging apps, conversion or editing, and a credential proves provenance, not accuracy.

4. Similarity, safety systems and contact

Output may resemble other users' output or existing designs, products or copyrighted works. You own your Output under the Terms, but you are responsible for checking it and for any rights you need before publishing it. Automated file checks and provider safety systems may block lawful material or miss harmful material, and provider outages or model changes can delay or prevent output; when an action fails before delivering, its reserved credits are released. Questions about AI generation or labeling: legal@kansumi.com.