Systems for city halls and public agencies
AI in city hall systems without exposing citizens.
Requests, case files, protocols and ombudsman messages carry the CPF, name, address and phone number of the people who use public services. Rota removes this data before AI and before publication, and your system keeps proof of the processing.
Why this matters in the public sector
The processing of personal data by public authorities and the use of artificial intelligence are among the inspection priorities of the ANPD, Brazil's National Data Protection Authority, for 2026 and 2027.
Transparency requires the publication of acts and documents. Brazil's General Data Protection Law (LGPD) requires protection of the personal data in them. The system must do both in the same flow.
The company that supplies the system to the agency is also accountable for the processing that it performs. A cleaning layer with records helps show what was done with each piece of data.
How it works
- The document arrives Your system sends the text of the request, decision or message to the cleaning route. PDFs go through the extraction route.
- Rota removes the personal data CPF and CNPJ numbers, names, emails, phone numbers, addresses, and protocol and registration numbers become placeholders or are removed, according to the agency's policy.
- AI helps with the task The model summarizes the request, classifies the subject or prepares a draft, without seeing who the citizen is.
- The system publishes or replies Your system completes the reply with the data that it needs and publishes the clean version when required.
Example with fictitious data
Before Rota
Protocolo 2026/004512. A munícipe Luciana Ferreira Gomes, CPF 529.982.247-25, moradora da Rua Sete de Setembro, 1200, pede a poda de uma árvore em frente ao imóvel. Contato: (41) 90000-2468.
After Rota
[REMOVIDO]. A munícipe [REMOVIDO], CPF [REMOVIDO], moradora da [REMOVIDO], pede a poda de uma árvore em frente ao imóvel. Contato: [REMOVIDO].
Fictitious text in Portuguese, processed by the API engine on 6 October 2026, with the redaction policy.
How to integrate
Use the redaction policy when the text will be published: the data is removed and shows as [REMOVIDO], as in the example above. Use placeholders when the AI needs to tell the people in the text apart.
The request policy can only make the organization's policy stricter: a request can ask for redaction, but it cannot relax a block rule.
POST /v1/privacy/clean
curl https://rota-nacional.ia.br/v1/privacy/clean \
-H "Authorization: Bearer $ROTA_API_KEY" \
-H "Content-Type: application/json" \
-d '{"text": "A munícipe Luciana Ferreira Gomes, CPF 529.982.247-25, pede a poda de uma árvore.", "policy": "redact"}'Good to know
- The agency sets the publication rule, and Rota applies the chosen policy: placeholder, redaction or block.
- The usual review still happens before publication.
FAQ
Frequently asked questions
Direct answers about this use. Did not find yours? Talk to the team.
Who integrates: the agency or the system supplier?
Usually the system supplier. The supplier connects the Rota API to the existing flow, and the agency sets the policy: placeholder, redaction or block.
Can we publish the document without the CPF and keep the rest?
Yes. With the redaction policy, the CPF, the name and the contact details are removed, and the rest stays the same.
Does Rota store the documents?
The PDF file is not stored, and standard records keep only categories and counts. When auditing is active, the input and the output stay encrypted for the defined period, with restricted access.