AI & Data Use

Searches go to the live source

A DGSI search is sent to dgsi.pt and a legislation or official-publication search is sent to Diário da República. Lauyer fetches and formats the response for your MCP client. It does not maintain a shadow copy of the full legal corpus.

The upstream source can see the request coming from Lauyer’s Cloudflare infrastructure and may retain its own technical logs. Your MCP client also sees the query and result.

AI summaries leave Lauyer

A summary request sends the relevant legal text and summary instructions to OpenRouter, which routes the request to an underlying model provider currently using Google Gemini. This is separate from an ordinary search or fetch.

OpenRouter says it does not train its own models on API inputs or outputs. Underlying model providers can have different retention and training terms. Do not submit confidential or sensitive material unless those provider terms and your duties permit it.

What Lauyer stores about AI use

Lauyer stores your account ID, the summary tool and source entity ID, model ID, input and output token counts, estimated cost, and timestamp. This allows quota enforcement, cost control, fraud prevention, and account support.

The usage ledger does not intentionally contain the full document, prompt text, or generated summary. Ordinary infrastructure and provider logs may still contain technical request information under their own retention rules.

AI is not the legal source

Generated text can be fluent and wrong. It may invent citations or legal effects, omit exceptions, rely on a headnote, or miss later amendments and decisions. Lauyer’s prompts reduce these risks but cannot eliminate them.

Open the DGSI or Diário da República link, read the controlling text, confirm status and date, and apply qualified human review before relying on a result.

Personal and confidential information

Official decisions may contain personal data even when lawfully public. Search and AI use must still respect purpose limitation, professional secrecy, employment duties, court restrictions, and data-protection law.

Do not add client secrets, privileged communications, unpublished case files, employee records, special-category data, criminal-offence data, or unrelated third-party identifiers to a query unless external processing is lawful, necessary, and protected.

Your MCP client has its own data path

Claude, ChatGPT, Cursor, Codex, or another client may store your conversation, tool-call parameters, authorizations, and results. Workspace administrators may also have retention or audit access. Review that client’s settings and contract before using Lauyer for professional work.

Human review checklist

• Open and read the original source.
• Confirm court, process, date, publication, version, and procedural status.
• Check amendments, revocations, rectifications, appeals, and later authority.
• Separate source text from AI interpretation.
• Remove unnecessary personal or confidential information before sharing.
• Have a qualified person review consequential work.

Questions and rights

For the complete legal treatment, read the Privacy Policy and Terms of Service. Questions and data-rights requests may be sent to privacy@lauyer.org.