AI use notice

AI can suggest functions, viewpoints, matrix decisions and test cases from supplied requirements and context. Unless a feature expressly says otherwise, it does not run your software, inspect a production system, execute the proposed cases or verify the truth of your requirements. Labels, structured JSON and confident language are not guarantees of correctness.

Output can be incomplete, inconsistent, biased, outdated or fabricated and can miss important failure modes. Review assumptions, coverage, expected results and source material before relying on, exporting or sharing it. Qualified people remain responsible for release, security, compliance and other consequential decisions. Do not use the service as the sole basis for safety-critical decisions.

When you request AI assistance, the application sends the requirements, existing design material and instructions needed for that action from its server to OpenAI. This can include relevant project context beyond the text currently visible in an input box. Hosting and database providers process the request or stored content as described in the Privacy notice. The current integration calls OpenAI directly from our hosting environment; Lovable is not the current AI gateway.

Submit only material you are authorised to disclose for this purpose. Use synthetic examples or remove unnecessary personal and confidential information. Do not submit credentials, secrets or regulated data requiring controls that we have not agreed to provide. Contact us before using AI for material subject to special contractual restrictions.

PDF and screenshot extraction uses AI to read text and visual content. Check the editable draft against the original before adding and saving it. OCR, page references and interpretations of interface behavior can be incorrect or incomplete. Source labels are navigation aids, not proof of verification. Original files are stored privately and can be previewed, downloaded or removed. Unchanged inputs reuse their extracted results. Saved requirements and usage records are retained separately as described in our Privacy notice. Imports temporarily reserve 60 available credits and settle using actual input and output token usage at the published rates. This reservation is an estimate, not a price cap; images and PDF page processing may use more tokens than plain text.

The current OpenAI request uses store: false. This is not a promise of zero retention: provider abuse-monitoring, security, legal and other applicable processing can still apply. OpenAI states that API data is not used for model training by default unless the customer opts in; applicable provider terms and account settings govern. We do not promise that a special Zero Data Retention arrangement applies to this account.

Kanten stores relevant results and usage records to support saved work, retries, credit accounting and service operation. A short cache or retry window is not a deletion schedule for database records. See the Privacy notice for retention and deletion requests. Kanten does not use your project content to train its own general-purpose AI model.

AI availability depends on account eligibility, remaining credits, service limits and provider availability. Validly completed output may consume credits even if it is not useful to you; a suggestion is not sold with a guarantee of suitability. Reservations and final charges can differ. Report a suspected duplicate charge or a failed action that consumed credits through Contact us so it can be investigated under the Billing policy.

AI output may resemble material generated for other users and may contain third-party material or errors. We do not guarantee exclusivity, copyright protection or freedom from third-party claims. Review output rights before publication. Content and output rights are addressed in the Terms. Models, prompts and capabilities may change; material changes to data processing will be addressed through updated notices and any consent required by law.