Insights

AI and legal ethics: the lawyer's duties

3 min read · Editorial oversight: Avv. Federico Papa

The integration of AI into the legal profession requires full compliance with the principles of the Code of Conduct. Although there are no specific rules yet, the duties of competence and personal performance require lawyers to maintain the primacy of human intellectual input, avoiding the complete delegation of legal opinions or court filings to algorithms.

In brief

Integrating AI into the legal profession requires compliance with the Code of Conduct. Articles 14 and 50 mandate duties of competence and truthfulness, requiring human supervision to prevent legal hallucinations. EU Regulation 2024/1689 (AI Act), effective August 2, 2026, imposes fines up to 15 million euros on deployers for transparency violations. Uploading personal data to generic models violates confidentiality and GDPR. The Rome Bar Association Vademecum recommends double verification and client disclosure. The National Bar Council has included AI in mandatory continuing legal education.

1.
The duty of competence and the Code of Conduct
The Code of Conduct imposes the duty of competence (Art. 14). The use of AI is considered a mere supporting tool: professionals must possess the preparation needed to supervise technological outputs, ensuring that human intellectual input remains central in every filed document.
2.
Duty of truthfulness and legal hallucinations
Case law has already highlighted the risks of non-existent legal citations: international courts have imposed financial penalties for the negligent use of AI. Lawyers are liable for violating the duties of truthfulness (Art. 50) and competence (Art. 14) under the Code of Conduct if they fail to verify every single source cited by AI.
3.
Confidentiality and AI Act sanctions
As of August 2, 2026, EU Regulation 2024/1689 (AI Act) imposes fines of up to 15 million euros (or 3% of global turnover) on deployers who violate transparency obligations. Law firms must avoid uploading personal or confidential data into unprotected general models, selecting GDPR-compliant European infrastructures such as edit.legal.
Apply this research directly with edit.legal
Legal research and drafting with citations checked against official databases. Try edit.legal for free, no credit card.
Try edit.legal
4.
The choice of tool as an act of diligence
Professional diligence requires selecting specialized tools that query verified official sources. While general models may fabricate precedents, edit.legal analyzes over 1 million documents, including legislation, Supreme Court rulings, and the Official Gazette, ensuring accurate and verified citations.
5.
Duty of disclosure and transparency to the client
Providing specific disclosure on the use of AI systems is a recommended best practice to ensure maximum transparency in executing the client mandate. Although not explicitly mandated by the Code of Conduct, this step supports compliance with the general duties of fairness and loyalty.
6.
Minimum law firm policy in three rules
The Rome Bar Association Vademecum suggests three core rules: the Data Shield (a prohibition against uploading identifying data to generic systems), double verification against official databases, and client transparency via written disclosure regarding the use of AI systems.
7.
Technology adoption and mandatory training
The CENSIS Report reveals steady growth in AI adoption among legal professionals, while the National Bar Council (CNF) has included the subject in mandatory continuing legal education. Internal metrics from edit.legal confirm that user queries increasingly focus on the reliability of automated legal sources.

Editorial oversight: Avv. Federico Papa·ICAM

Frequently asked questions

What sanctions do I risk if AI fabricates a court ruling?

In addition to civil liability for frivolous litigation under Art. 96 of the Italian Code of Civil Procedure, lawyers face disciplinary sanctions for violating the duties of truthfulness and competence. Foreign court rulings have already imposed financial penalties for failing to verify cited sources.

Can I enter client names into ChatGPT or Claude?

No, unless using an enterprise version that contractually guarantees data protection and a Training-Free clause. From an ethical perspective, uploading identifying data into generic systems violates confidentiality duties and conflicts with AI Act rules effective from August 2026.

Must the use of AI be communicated to the client?

Although not explicitly required by the Code of Conduct, informing the client about the use of AI tools is a recommended best practice to fulfill general duties of professional fairness and transparency.