We use cookies to personalise the website and offer you the greatest added value. They are, among other purposes, used to analyse visitor usage in order to improve the website for you. By using this website, you agree to their use. Further information can be found in our data privacy statement.



Legislative Decree No. 160/2026: New Rules in Italy on Criminal Liability, Corporate Liability and Claims for Damages

​​​​​​​​​​​​​​published on 18 September 2026 | reading time approx. 4 minutes​​​

​
Italian Legislative Decree No. 160 of 9 September 2026, which enters into force on 30 September 2026, completes the process of bringing Italian law into line with the EU AI Act and directly affects liability arising from the development and use of artificial intelligence systems. For companies, the change is significant: technical safety, human oversight and documentation are no longer merely compliance safeguards, but also become relevant to criminal liability, corporate liability under Legislative Decree No. 231/2001 and damages claims.

New criminal offences and corporate liability under Legislative Decree No. 231/2001​​

Article 12 introduces new Article 437-bis into the Italian Criminal Code, addressing high-risk AI systems. Failure to implement the technical safety measures and human oversight safeguards required to prevent malfunctions or alterations is punishable where it creates a risk to life or to public or individual safety. The penalty is imprisonment for one to five years, increasing to two to eight years where State security is endangered. Unlawfully altering the system is punishable by imprisonment for two to six years where it creates a risk to life or to public or individual safety, and by imprisonment for three to ten years where State security is endangered; in the circumstances specified by the provision, gross negligence may also give rise to liability. The provision also applies to professional deployers that intentionally fail to implement human oversight safeguards.

Article 15 also adds new Article 25-vicies to Legislative Decree No. 231/2001. The offence under Article 437-bis of the Criminal Code carries a pecuniary sanction ranging from 600 to 1,000 quotas; the unlawful dissemination of content generated or altered using AI systems under Article 612-quater of the Criminal Code carries a sanction ranging from 200 to 700 quotas. Disqualification sanctions also apply. This does not, in itself, require an across-the-board update of 231 Models: companies must assess their actual exposure to the new offences and, where necessary, update their risk assessment, preventive protocols and reporting flows to the Supervisory Body.

Civil litigation and access to evidence​​

The Decree introduces specific evidentiary mechanisms for all contractual and non-contractual claims for damages connected with the use of AI systems, irrespective of their risk classification. Where the claim appears plausible, the court may order the opposing party or a third party to produce relevant evidence concerning the operation of the system. The evidence expressly identified includes: logs, the risk management system, technical documentation, system parameters and human oversight arrangements.
An unjustified failure to comply with a production order may allow the court to draw adverse inferences.

Where the failure concerns the documentation specifically identified in Article 17, the court, having considered the other evidence, will deem the facts alleged by the claimant to be admitted. A non-compliant third party may be fined between EUR 1,500 and EUR 10,000. Any production order must nevertheless be necessary and proportionate and must balance access to evidence against the protection of trade secrets and confidential information.

Article 18 also introduces a rebuttable presumption as to causation: where damage results from a breach of one or more obligations under the AI Act, the causal link between the breach and the damage is presumed unless rebutted. This is not a general presumption of liability; rather, it places on the defendant the burden of rebutting causation. A company’s ability to reconstruct how the system operated, the human interventions performed and the measures implemented will therefore be critical. Compliance with the AI Act, even where certified, does not automatically preclude liability in damages.

As regards insurance, before commencing proceedings the injured party may ask whether the party considered liable has civil liability insurance. Within thirty days, the recipient must disclose whether a policy exists, provide its details and identify the insurer; the court may draw adverse inferences from an omitted or incomplete response. The injured party is also granted a direct right of action against the insurer, up to the applicable policy limit.

Key priorities for Italian companies​

  • map high-risk AI systems and assess exposure to the new criminal offences;
  • test and document technical measures, human oversight, logs, risk management and technical documentation;
  • incorporate Article 25-vicies into the 231 risk assessment and, where necessary, update the 231 Model and related protocols;
  • establish rules for preserving and producing evidence while protecting trade secrets and confidential information;
  • review the contractual allocation of risk, insurance coverage and procedures for responding to requests for policy information.

Tech & Data Bites

author

Contact Person Picture

Nadia Martini

Attorney at Law (Italy)

Partner

+39 02 6328 841

Send inquiry

Profile

Skip Ribbon Commands
Skip to main content
Deutschland Weltweit Search Menu