AI for business.
Privacy before inference.

The rn.* namespace groups capabilities in development for chat, embeddings, audio, OCR, documents and images. Endpoint examples are an integration preview and may change. Current transcription and extraction remain available through their documented routes; this does not mean the new RN execution is released. Contact the team before planning production use.

It identifies personal data in submitted content and applies your organization's policy: placeholders, removal or blocking. The dashboard records categories and usage. Detection has limitations; results need to be evaluated for the process in which they will be used.

No. You can sign in with Google and test text in your browser. To apply sanitization within an automated organizational process, your technical team needs to integrate the API. Extensions and ready-made connectors are still planned.

The available version uses patterns for email, phone numbers, Brazilian identifiers and other categories. Numeric CPF and CNPJ identifiers undergo check-digit validation. General name recognition remains under evaluation; names currently depend on a configured list. See the rules and limitations in the engineering documentation.

On the engineering page, we publish results from 46 synthetic texts, remaining errors, annotated data, predictions and a scoring script. The same page maps papers and standards to the techniques used. This is experimental evidence, without certification or guaranteed production coverage.

No. These are ten fictional Portuguese texts with precomputed results from the expanded detection engine under evaluation. You can compare the examples without submitting data. This version is not yet serving customers; the results do not guarantee the same coverage on other texts.

The trial lasts 30 days, with a starter quota and no card required. Continued use after that period requires commercial approval. To evaluate use in your organization, talk to the team about volume, integration and commercial terms.

Registration is open with Google, anti-abuse verification and acceptance of the terms. If you have set up a passkey, you can also use it to sign in. The dashboard trial uses the operational rules; expanded detection shown in the examples and benchmarks remains under evaluation.

Registration explains the terms and encrypted retention of input and output. When auditing is enabled, this content may be retained for the specified period with restricted access. Standard usage records show metadata, such as categories and counts, without raw text.

Operations and curation are Brazilian, but the current service does not guarantee that all data and inference remain exclusively in Brazil. Dedicated deployment, jurisdiction-specific residency and offline operation require a specific project and validation.

Detection reduces exposure but cannot eliminate every risk. Monitoring policies do not sanitize data; OCR is in beta. Compliance also depends on the organization's configuration and processes.

No. Your team can use the API solely to sanitize data and send the result to its chosen destination. Rota also provides access to AI models if that suits your integration.