CORPORATE CRIMINAL LIABILITY FOR AI-RELATED CRIMES:
POSSIBLE LEGAL TECHNIQUES AND OBSTACLES
By Vincenzo Mongillo*
Abstract
The use of Artificial Intelligence in corporations can lead to significant benefits in terms of
strengthening the criminal compliance system, but it can also entail significant risks when the
actions of those digital tools could cause danger or damage that may result in a criminal offence.
This special report will focus on the latter research perspective, first exploring the possibility that
an entity may be held accountable for AI-related crimes by looking at the current regulatory
framework, and then investigating possible techniques to make corporations criminally liable in
this field, as well as potential counter arguments to such a regulatory objective.
1 Introduction
Artificial Intelligence is an attractive research field for corporate criminal liability schol-
ars for at least two reasons. The first reason is that new technologies, such as Artificial
Intelligence and Blockchain, are likely to radically change criminal compliance and the
design of internal control systems in corporations. In the general context of what we
might call the ‘RegTech World’, by using these sophisticated IT systems corporations
may detect red flags that they would not otherwise be able to identify through traditional
analysis. Additionally, they can build transparent and reliable decision-making systems
(e.g., through a blockchain architecture) in which it is thus more difficult to conceal un-
lawful activities.1 The second reason is that the use of an Artificial Intelligence system in
the context of a corporation can cause damage or danger that may result in a criminal
offence. In this regard, the key question is whether the legal entity can – de iure condito –
or should – de iure condendo – be held criminally liable.
* Full Professor of Criminal Law at University of Rome ‘Unitelma Sapienza’, vincenzo.mongillo@unitel-
masapienza.it.
1 On this topic, also with respect to the use of technology not only as a tool for crime prevention but also
as an instrument for the commission of criminal offences, see Lorenzo Picotti, ‘New Technologies as Tools
for and Means Against Crime: Substantive Aspects’ [2020] 2 RIDP 183. For a recent overview and a
broader literature review with regard to this issue, see Emanuele Birritteri, ‘Corporate Criminal Liability
and New Technologies: Digital Compliance Strategies in the Fight Against Economic Crimes’, Proceed-
ings of International Conference organised by the OECD Presence in Albania in cooperation with the
Faculty of Law of Tirana University and with the participation of CEMAS (Centre for Scientific Research
and Cooperation with Eurasia, the Mediterranean and Sub-Saharan Africa) of Sapienza University of
Rome: The Role of Technology in Preventing and Combating Organized Crime, Financial Crimes and Corruption
(forthcoming). See also Vincenzo Mongillo, ‘Presente e futuro della compliance penale’ available at <
https://www.sistemapenale.it >, accessed November 2022; Antonio Gullo, ‘Compliance’ in Carlo
Piergallini, Grazia Mannozzi, Carlo Sotis, Chiara Perini, Marco Scoletta, Federico Consulich (eds.), Studi
in onore di Carlo Enrico Paliero (Giuffrè 2022); and Rossella Sabia, ‘Artificial Intelligence and Environmental
Criminal Compliance’ [2020] 1 RIDP 179 ff., and Paola Severino, ‘The Importance of Corporate
Compliance in the Digital Era’ [2021] 2 RIDP 435 ff.
77
, This special report will focus on the latter perspective, first exploring the possibility that
an organisation may be held accountable for AI-related crimes by looking at the current
regulatory framework,2 and then investigating possible techniques to make corporations
criminally liable in this field as well as potential counter arguments to such a regulatory
aim.3
2 Corporate Criminal Liability Rules Internationally and AI-related Crimes: The
Current Scenario
When it comes to the question of whether a legal person can be held liable for an AI-
related crime, a distinction must be drawn between non-autonomous AI systems and AI
systems that can make totally autonomous decisions, which the programmers or users
themselves are unable to foresee in whole or in part.4
Starting from the first case, non-autonomous systems do not seem to entail any issues over
and above those traditionally arising in criminal matters (and in corporate criminal lia-
bility law) with regard to product liability. These AI systems merely execute the instruc-
tions received from the ‘human’ programmer or, in any case, operate under the direct
control of a human being. Consequently, the ‘ordinary’ enforcement tools of criminal
law can be triggered.5 In fact, both where, on the one hand, the instrument is used with
the specific purpose of committing a crime and, on the other hand, the use of the system
causes unintentional (i.e., negligent) harm that may result in the occurrence of a crime,6
liability may be allocated according to the usual rules. As a consequence, it is possible to
ascribe liability to, as the case may be, the producer, distributor or user of the system,7
as well as the legal person in accordance with the different ‘rules of engagement’ pro-
vided for under the various corporate criminal liability rules. The different mechanisms
range from those based on strict liability or vicarious liability, to those focused on organ-
isational fault or failure to prevent specific offences.8
Obviously, liability issues may arise also in such cases, both in terms of individual lia-
bility and corporate liability. From the perspective of corporate criminal liability, for in-
stance, consider the case in which the use of an AI system – e.g., a self-driving car under
the supervision of a human driver – leads to the commission of manslaughter and the
2 See Section 2.
3 See Section 3.
4 For a general overview see Susanne Beck, ‘The Problem of Ascribing Legal Responsibility in the Case of
Robotics’ [2016] 31 AI & Soc. 473 ff.
5 On this issue see also Beatrice Panattoni, ‘AI and Criminal Law: the Myth of ‘Control’ in a Data-Driven
Society’ [2021] 1 RIDP 125 ff.
6 A review of the possible uses of AI systems in the perpetration of crimes was made by Matthew Cald-
well et al., ‘AI-enabled future crime’ [2020] 9 Crime Sci 14.
7 In the Italian literature see Carlo Piergallini, ‘Intelligenza artificiale: da ‘mezzo’ ad ‘autore’ del reato? ‘
[2020] 4 Rivista italiana di diritto e procedura penale 1753.
8 On this topic see also Federico Mazzacuva, ‘The Impact of AI on Corporate Criminal Liability: Algorith-
mic Misconduct in the Prism of Derivative and Holistic Theories’ [2021] 1 RIDP 143 ff.
78
