Regulatory sandboxes - Liability aspects of a new legislative method
(2023) JURM02 20231Department of Law
- Abstract
- In this paper, I investigate how the EU plans to address AI in the area of non-contractual liability. I present two proposed directives and discuss how these supplement each other to create a liability regime that remains potent as the cases to which it is applied grow increasingly complicated and condi-tions harder to prove. Secondly, I address the concept of regulatory sand-boxes, its potential role in the legislative processes of tomorrow and its vir-tues from the perspective of AI developers. A common element of worry, I also discuss which level of liability is appropriate to apply in a regulatory sandbox and make two main conclusions. Firstly, my interpretation of the AI Act is that regulatory sandboxes might indeed offer some... (More)
- In this paper, I investigate how the EU plans to address AI in the area of non-contractual liability. I present two proposed directives and discuss how these supplement each other to create a liability regime that remains potent as the cases to which it is applied grow increasingly complicated and condi-tions harder to prove. Secondly, I address the concept of regulatory sand-boxes, its potential role in the legislative processes of tomorrow and its vir-tues from the perspective of AI developers. A common element of worry, I also discuss which level of liability is appropriate to apply in a regulatory sandbox and make two main conclusions. Firstly, my interpretation of the AI Act is that regulatory sandboxes might indeed offer some significant advantages to prospective participants in terms of liability, by creating an opportunity to test AI inventions in real world conditions without being subject to product liability. This opportunity may however change, since the mention of testing in real world conditions, within the boundaries of a sand-box, does not exist in all versions of the proposal. It remains to be seen if it ends up a part of the final act. A more important function of regulatory sandboxes, however, is that they create an atmosphere where multiple ac-tors, from both the private and the public sector, can share the burden of liability. I argue that other parties than the prospective provider of an AI system could, and should, be held liable for damage to the extent of their participation in the sandbox procedure. With this approach, a sandbox host could be held liable for things such as negligent supervision and faulty ad-vice and a testing facility for negligent planning and execution of tests. Sec-ondly, I believe that, while it may be appropriate for public authorities to exercise their discretionary powers to remove regulatory obstacles from the development of technology, no such discretion does, or should exist in the realm of private, non-contractual liability law, since the victims of such measures would always be individuals. (Less)
- Abstract (Swedish)
- I denna uppsats undersöker jag mötet mellan AI och utomobligatorisk ska-deståndsrätt ur ett EU-perspektiv. Jag presenterar två föreslagna direktiv och diskuterar hur dessa kompletterar varandra för att skapa en lagstiftning som är lika effektiv i tekniskt komplicerade fall där rekvisiten är svårare att uppfylla. Jag undersöker också konceptet regulatorisk sandlåda, dess förde-lar ur ett utvecklarperspektiv och dess vikt i morgondagens lagstiftnings-processer. En viktig fråga som diskuteras i detta sammanhang är tidigare kritik mot det faktum att en regulatorisk sandlåda inte ger deltagarna någon civilrättslig ansvarslättnad. I denna uppsats undersöker jag i vilken mån kritiken fortfarande är rättfärdigad. Min uppfattning är att en... (More)
- I denna uppsats undersöker jag mötet mellan AI och utomobligatorisk ska-deståndsrätt ur ett EU-perspektiv. Jag presenterar två föreslagna direktiv och diskuterar hur dessa kompletterar varandra för att skapa en lagstiftning som är lika effektiv i tekniskt komplicerade fall där rekvisiten är svårare att uppfylla. Jag undersöker också konceptet regulatorisk sandlåda, dess förde-lar ur ett utvecklarperspektiv och dess vikt i morgondagens lagstiftnings-processer. En viktig fråga som diskuteras i detta sammanhang är tidigare kritik mot det faktum att en regulatorisk sandlåda inte ger deltagarna någon civilrättslig ansvarslättnad. I denna uppsats undersöker jag i vilken mån kritiken fortfarande är rättfärdigad. Min uppfattning är att en regulatorisk sandlåda skulle kunna erbjuda deltagare ansvarslättnad, genom att erbjuda möjlighet att testa ny teknologi i verkliga scenarion, men utan att Produkt-ansvarsdirektivet är tillämpligt. Det är osäkert om detta kommer vara fallet i den slutgiltiga versionen av AI Act, eftersom den formulering som enligt mig tillåter en sådan tolkning inte finns med i alla versioner. En viktigare egenskap hos en regulatorisk sandlåda i AI Act är dock att den erbjuder en miljö där skadeståndsansvar kan delas upp mellan flera olika aktörer från såväl privat som offentlig sektor. Inte bara utvecklare, utan andra medver-kande bör därför också hållas ansvariga i den utsträckning de bidrar till projektet. Ansvarig myndighet för en regulatorisk sandlåda skulle på detta sätt kunna hållas ansvarig för oaktsamhet i tillsyn och vägledning, liksom en testanläggning skulle kunna hållas ansvarig för att inte ha planerat och genomfört experiment på ett säkert sätt. I enlighet med detta anser jag det olämpligt att lätta på det civilrättsliga ansvaret eftersom en sådan åtgärd skulle placera en orimligt stor börda på enskilda personer. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9133649
- author
- Olsson, Måns LU
- supervisor
- organization
- course
- JURM02 20231
- year
- 2023
- type
- H3 - Professional qualifications (4 Years - )
- subject
- language
- English
- id
- 9133649
- date added to LUP
- 2023-09-08 10:48:50
- date last changed
- 2023-09-08 10:48:50
@misc{9133649,
abstract = {{In this paper, I investigate how the EU plans to address AI in the area of non-contractual liability. I present two proposed directives and discuss how these supplement each other to create a liability regime that remains potent as the cases to which it is applied grow increasingly complicated and condi-tions harder to prove. Secondly, I address the concept of regulatory sand-boxes, its potential role in the legislative processes of tomorrow and its vir-tues from the perspective of AI developers. A common element of worry, I also discuss which level of liability is appropriate to apply in a regulatory sandbox and make two main conclusions. Firstly, my interpretation of the AI Act is that regulatory sandboxes might indeed offer some significant advantages to prospective participants in terms of liability, by creating an opportunity to test AI inventions in real world conditions without being subject to product liability. This opportunity may however change, since the mention of testing in real world conditions, within the boundaries of a sand-box, does not exist in all versions of the proposal. It remains to be seen if it ends up a part of the final act. A more important function of regulatory sandboxes, however, is that they create an atmosphere where multiple ac-tors, from both the private and the public sector, can share the burden of liability. I argue that other parties than the prospective provider of an AI system could, and should, be held liable for damage to the extent of their participation in the sandbox procedure. With this approach, a sandbox host could be held liable for things such as negligent supervision and faulty ad-vice and a testing facility for negligent planning and execution of tests. Sec-ondly, I believe that, while it may be appropriate for public authorities to exercise their discretionary powers to remove regulatory obstacles from the development of technology, no such discretion does, or should exist in the realm of private, non-contractual liability law, since the victims of such measures would always be individuals.}},
author = {{Olsson, Måns}},
language = {{eng}},
note = {{Student Paper}},
title = {{Regulatory sandboxes - Liability aspects of a new legislative method}},
year = {{2023}},
}