About the Directive
As reported by Euractiv, the AI Liability Directive is due to be published on September 28 and is intended to complement the Artificial Intelligence Act (read more about the Act here). It is expected to provide a legal basis for claiming damages when an entity fails to fulfill its obligations under the AI law.
The Directive’s aim is to apply to non-contractual civil law claims for damages caused by an artificial intelligence system in fault-based liability regimes, that is, where a person may be liable for a specific act or omission.
The causal link between the failure to comply with AI obligations and the resulting damage can only be established by explaining the internal workings of artificial intelligence. The directive presupposes the existence of a causal relationship under certain circumstances.
When dealing with an artificial intelligence system that does not represent a particular level of risk, the presumption applies, but only if non-compliance with rules that could have prevented the damage is demonstrated and if the defendant is responsible for this non-compliance.
In the case of high-risk systems, the presumption applies against the provider if it failed to apply appropriate risk management measures, or the training data set did not meet the required quality criteria, as well as if the system does not meet transparency, accuracy, or cybersecurity criteria.
Other factors include a lack of adequate human oversight or negligence in immediately implementing necessary corrective measures.
Information request
The plaintiff will have the right to ask providers of high-risk AI systems to disclose the information they will have to keep as part of their obligations under the AI Act. The information is to include i.e. the datasets used to develop the AI system, or technical documentation.
The provider of the system will have the right to refuse the request, which can then be raised again through a lawsuit, where a judge will assess whether it is reasonable and necessary to sustain a claim for accidents involving AI.
If the recipient of the request refuses to comply with an order to disclose the information, the court will assume that it has failed to comply with its obligations, unless the defendant proves otherwise.
The role of the European Commission
What’s more, the directive stipulates that the European Commission will monitor incidents involving artificial intelligence systems, and within five years make proposals for new, additional measures when it assesses that they are needed.
The Member Countries will have two years to implement the directive into national law from its date of coming into effect.
Author: Agata Konieczna
Source: Euractive: Article: LEAK: Commission to propose rebuttable presumption for AI-related damages
Credits: Liability icons created by Freepik – Flaticon
Failure icons created by Iconjam – Flaticon
Greetings! Very useful advice within this article! Its the little changes that make the most significant changes. Thanks a lot for sharing!