Italy has taken a further step towards establishing a national framework for civil liability arising from the use of artificial intelligence systems. Legislative Decree No. 160 of 9 September 2026 implements part of the framework established by Law No. 132/2025 on Artificial Intelligence, introducing specific procedural rules for claims concerning damage caused through the use of AI systems.

‍
The Decree was published in the Italian Official Journal, General Series No. 214, on 15 September 2026, and will enter into force on 30 September 2026. Its civil-liability provisions are contained in Articles 16–20, under Chapter II of Title II, expressly entitled “Civil procedural instruments for compensation for damage caused by the use of artificial intelligence systems.”
The structure of the new framework is significant. Rather than establishing a comprehensive and autonomous regime of strict liability for AI, the Italian legislature has primarily intervened on the procedural and evidentiary dimension of civil litigation. The new rules are designed to facilitate the claimant’s access to evidence, modify the allocation of the burden of proof in specific circumstances, and facilitate access to compensation.

‍
1. Access to Evidence and Information Asymmetry
The first major innovation concerns access to evidence relating to the operation of AI systems.
Under Article 17, in actions for contractual or non-contractual damages caused through the use of an AI system, the court may order the defendant or a third party possessing the relevant information to disclose specifically relevant evidence concerning the operation of the system. The mechanism is available where the claimant presents facts and elements capable of making the claim sufficiently plausible.
The provision is therefore aimed at addressing the information asymmetry that may characterise AI-related litigation. Relevant evidence concerning the functioning of an AI system may be held primarily by the provider, deployer or another actor involved in the system’s operation. The claimant may consequently face substantial difficulties in establishing the factual basis of a claim without access to such information.
The Decree also establishes consequences for failure to comply with an order to disclose evidence. Where a party fails to comply without justification, the court may draw evidentiary inferences pursuant to Article 116 of the Italian Code of Civil Procedure. For the categories of documents specifically identified by the provision, the court may, after considering the other evidence available, treat the facts alleged by the claimant as admitted. Where the order is addressed to a third party, unjustified non-compliance may result in a monetary penalty ranging from EUR 1,500 to EUR 10,000.
Article 16 complements this mechanism by establishing the scope of the provisions and introducing an additional territorial forum for a natural person who has suffered damage, thereby strengthening access to judicial protection.

‍
2. A Rebuttable Presumption of Causation
The second central element is contained in Article 18, which introduces a specific rebuttable presumption of causation.
Where the damage results from the violation of one or more obligations imposed by the AI Act, the causal link between the violation and the damage is presumed, unless the contrary is proved.
The scope of this presumption must, however, be carefully defined. It concerns the causal relationship between the violation of an AI Act obligation and the damage. It does not eliminate the claimant’s burden of proving the underlying violation itself. Nor does the presumption extend to fault, which remains subject to proof by the claimant where fault is required under the applicable liability regime.
Article 18 should therefore not be understood as introducing automatic or strict liability for AI-related harm. Rather, it modifies the evidentiary structure of civil claims in a specific category of cases: once the relevant AI Act violation and the other necessary elements have been established, the causal link benefits from a rebuttable presumption.

‍
3. AI Act Compliance Does Not Automatically Exclude Civil Liability
A third important element concerns the relationship between regulatory compliance and civil liability.
Under Article 19, the fact that an AI system complies with the obligations established by the AI Act does not, in itself, exclude the civil liability of the defendant. The rule applies even where compliance has been certified.
This provision is important because it prevents AI Act compliance from operating as a general civil-law safe harbour. Compliance with the regulatory requirements of the AI Act and the existence of civil liability remain conceptually distinct questions. The former may be relevant to the assessment of a dispute, but it does not automatically determine the latter.
‍

4. Direct Action against the Insurer
The fourth mechanism is contained in Article 20, which grants the injured party a direct action against the insurer providing civil-liability coverage to the defendant for damage caused through the use of AI systems.
The provision adds another element to the new procedural framework. The Italian legislature is therefore not only addressing the evidentiary difficulties involved in proving AI-related harm, but also introducing a mechanism intended to facilitate the practical recovery of compensation where insurance coverage is available.
Taken together, Articles 17, 18 and 20 establish a combination of evidence disclosure, a rebuttable presumption of causation and direct recourse against the insurer. Article 19 complements these mechanisms by clarifying that regulatory compliance does not, by itself, exclude civil liability.

‍
5. Interaction with the EU Product Liability Framework
The new Italian provisions must also be considered alongside the evolving European framework on liability for defective products, in particular Directive (EU) 2024/2853 on liability for defective products.
The two regimes should not be conflated. The Italian Decree establishes procedural mechanisms applicable to civil claims concerning damage caused through the use of AI systems, whereas the Product Liability Directive establishes a specific no-fault liability regime for defective products. The Directive also contains its own rules on disclosure and the burden of proof. Under Article 10, for example, the claimant must generally prove the defectiveness of the product, the damage and the causal link, while the Directive establishes a series of presumptions concerning defectiveness and causation, including where the claimant faces excessive difficulties in proving those elements because of technical or scientific complexity.
The interaction between the two frameworks will therefore be particularly relevant where an AI system falls within the material scope of the Product Liability Directive. In such cases, the applicable liability regime and the corresponding evidentiary mechanisms will have to be identified carefully rather than treating the Italian Decree as a general replacement for the European product-liability framework.

‍
6. From Delegation to an Operative Procedural Framework
The adoption of Legislative Decree No. 160/2026 marks the transition from the delegation contained in Law No. 132/2025 to a concrete national framework governing civil proceedings involving AI-related damage.
Its approach is deliberately narrower than the creation of a comprehensive “AI liability code”. The legislature has instead introduced targeted mechanisms addressing some of the difficulties that make AI-related litigation particularly complex: access to technical evidence, causal uncertainty, and the practical recovery of compensation.
From 30 September 2026, these provisions will therefore form part of the Italian legal framework applicable to civil claims involving AI systems. Their practical significance will depend not only on judicial interpretation of Articles 16–20, but also on their interaction with the existing rules of Italian contractual and non-contractual liability, the AI Act and the European Product Liability Directive.

‍
The Italian model thus represents an important development in the emerging European architecture of AI liability: rather than creating a separate and self-contained liability regime, it strengthens the procedural position of injured parties while leaving the underlying categories of civil liability substantially in place.

‍