Professional Liability and AI Hallucinations: The Duty to Verify and Gross Negligence

1. 1.

   The facts: AI hallucinations in the Supreme Court

   The Criminal Court of Cassation, Section III, with ruling no. 23006 of June 22, 2026, addressed a case where an appeal contained entirely non-existent case law citations produced by AI software. The Court declared the appeal inadmissible, ordering the party to pay 5,000 euros to the Fine Fund. Similarly, in civil proceedings, the Court of Syracuse, Section II, with ruling no. 338 of February 20, 2026, sanctioned a plaintiff under Art. 96 c.p.c. for including fabricated legal principles in a brief, imposing a 2,000 euro fine plus legal costs.

2. 2.

   Legal interpretations: the duty of supervision

   2026 case law consolidated the human-in-the-loop principle. The use of AI is lawful, but the lack of supervision over outputs constitutes a breach of the duties of loyalty and probity under Art. 88 c.p.c. The Civil Court of Cassation, order no. 21279/2026, reiterates that sanctions for aggravated liability serve not only a compensatory but also a punitive function, aimed at discouraging the pollution of proceedings with false or unverified data.

3. 3.

   PROS: The benefits of strict sanctions

   The strict approach protects the reliance of the judge and opposing parties: the truthfulness of citations is a fundamental prerequisite for adversarial proceedings. Furthermore, sanctioning hallucinations ensures procedural efficiency, preventing judges from wasting time searching for non-existent rulings, and preserves the intellectual dignity of the legal profession, distinguishing the lawyer from a mere prompt executor.

4. 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.

   CONS: Risks of technophobia and training gaps

   There is a risk that overly severe sanctions may discourage technological innovation necessary for the international competitiveness of law firms. Often, the fault lies in a lack of specific training on digital tools, especially among junior associates. Additionally, the debate remains open regarding the liability of software companies that market legal-ready tools without sufficient guarantees against technical errors.

5. 5.

   Practical implications for law firms

   Art. 2236 c.c. no longer protects lawyers using AI without verification, as hallucinations are now a foreseeable risk. Firms must consider the risk of systematic reporting to Bar Councils for breaches of Arts. 11 and 12 of the Deontological Code. Furthermore, many professional indemnity insurance policies are introducing exclusion clauses for damages resulting from unsupervised automated systems.

6. 6.

   What to expect: the centrality of human control

   Human control in the decision-making process will become the minimum standard of diligence. Law firms will need to adopt internal mandatory validation protocols, such as manual double-checking of every legal citation. An increase in sanctions under Art. 96, paragraphs 3 and 4, c.p.c. is likely, serving as a mechanism to purge the judicial system of unsupervised automated drafting practices.

References: Cassazione Penale, Sez. III, sentenza n. 23006 del 22 giugno 2026Tribunale di Siracusa, Sez. II, sentenza n. 338 del 20 febbraio 2026Cassazione Civile, ordinanza n. 21279 del 2026Art. 96 c.p.c.Art. 88 c.p.c.Art. 2236 c.c.Art. 11 Codice Deontologico ForenseArt. 12 Codice Deontologico Forense

Editorial oversight: Avv. Federico Papa·ICAMUpdated 29 July 2026