AI Transparency
Last updated: 1 November 2026. This page explains how CIPRIAN-PAUL FRANDEȘ - CABINET DE AVOCAT (Frandeș Law) uses artificial intelligence, on this website and in its work for clients, with reference to Regulation (EU) 2024/1689 on artificial intelligence (the AI Act), as amended.
1. The chat assistant on this website
The chat assistant on frandes.law is an AI system. You are interacting with a machine, not with a lawyer or a member of staff. The chat window says so before you write your first message, as required by Article 50(1) of the AI Act, which applies from 2 August 2026.
The assistant:
- answers general questions about the firm, its practice areas and its services;
- can search the CAEN classification;
- can pass on your request for a consultation to our inbox, if you ask for one and give your contact details.
The assistant does not give legal advice, takes no decisions about you and may make mistakes. Its replies are generated by a model provided by Mistral AI (France). If Mistral rejects a request, an equivalent model from Microsoft (Azure OpenAI Service) or Google (Vertex AI) is used, processed in the EU. How your messages are processed is explained in our Privacy Notice.
2. AI in our work for clients
The firm uses an internal AI system, built in-house, which relies on language models from the providers listed on our Processors page. We use it to:
- draft and review legal documents, such as contracts, corporate resolutions, powers of attorney and notices;
- research legislation and case law, checking each reference against official sources;
- extract and cross-check information from public registers (ONRC, ANAF);
- calculate procedural and legal deadlines.
Names and personal identifiers from client files are pseudonymised before text is sent to an AI model, as explained on the Processors page.
The lawyer decides. The system takes no decisions on its own. Every output is a draft that the lawyer reviews, corrects where needed and approves before it is sent to a client or to a third party. The lawyer remains professionally responsible for all legal services provided.
3. How the AI Act applies
In our assessment, the uses described on this page are not high-risk uses under Article 6 and Annex III of the AI Act. Point 8(a) of Annex III covers AI systems used by or on behalf of judicial authorities, or in alternative dispute resolution, to research and interpret facts and the law and to apply the law. It does not cover the use of AI by a law firm to assist its own lawyers. Our system is also not a safety component of a product under Annex I.
The obligations that do apply to us are:
- ensuring that the people who use AI systems on our behalf have a sufficient level of AI literacy (Article 4, applicable since 2 February 2025);
- informing people when they interact with an AI system, as we do for the website chat (Article 50(1));
- disclosing AI-generated text published to inform the public on matters of public interest, unless the text has been reviewed by a person who holds editorial responsibility for it (Article 50(4)).
We will review this assessment if we change how we use AI or if the law changes.
4. Articles published on this website
Articles in the Insights section are published under the editorial responsibility of the firm. Where AI is used to help prepare an article, a lawyer of the firm reviews and edits the text before it is published.
5. Your choices as a client
In addition to what the law requires, we make the following commitments to our clients:
- Information: you can ask us at any time whether and how AI was used in your matter.
- Work without AI: you can ask that your documents be prepared without AI assistance. This does not affect the quality or the price of our services.
- Human review: you can ask the lawyer to review again, without AI assistance, any material prepared with AI assistance and to give you an independent opinion.
You also have the right of access to your personal data under Article 15 GDPR, as explained in our Privacy Notice. Requests go to office@frandes.law.
6. Known limitations
- Errors: language models can produce inaccurate information, references to provisions that do not exist or wrong legal interpretations. The lawyer checks every legally relevant output before it is used.
- Time limit of the training data: models are trained on data up to a certain date and do not reflect later changes in the law on their own. The lawyer checks every legislative reference against official sources (legislatie.just.ro, Monitorul Oficial).
- No independent legal judgment: AI is a working tool for the lawyer. It does not replace professional judgment, bears no professional responsibility and cannot weigh the circumstances of a case the way an experienced lawyer does.
Contact
For any question about our use of AI, write to office@frandes.law. We update this page when our use of AI or the applicable law changes.