kom (2022) 0496 - Ingen titel

Europaudvalget 2022
KOM (2022) 0496
Offentligt

EUROPEAN

COMMISSION

Brussels, 28.9.2022

SWD(2022) 318 final

COMMISSION STAFF WORKING DOCUMENT

AI Subsidiarity Grid

Accompanying the document

Proposal for a Directive of the European Parliament and of the Council

on adapting non contractual civil liability rules to artificial intelligence

(AI Liability Directive)

{COM(2022) 496 final} - {SEC(2022) 344 final} - {SWD(2022) 319 final} -

{SWD(2022) 320 final}

kom (2022) 0496 - Ingen titel

Subsidiarity Grid

1. Can the Union act? What is the legal basis and competence of the Unions’ intended action?

1.1 Which article(s) of the Treaty are used to support the legislative proposal or policy initiative?

The (proposal for a) Directive on adapting non-contractual civil liability rules to artificial intelligence is

based on Article 114 of the Treaty on the Functioning of the European Union (TFEU), which regulates

the approximation of the provisions laid down by law, regulation or administrative action in Member

States having as their object the establishment and functioning of the internal market. The Directive

harmonises targeted aspects of the Member States’ existing civil liability rules applicable to AI-

systems, in order to improve the conditions for the functioning of the internal market in AI-enabled

products and services. The choice of Article 114 TFEU as a legal basis is also supported by the EP

which has called twice upon the Commission to use this legal basis for a legislative proposal.

1.2 Is the Union competence represented by this Treaty article exclusive, shared or supporting in

nature?

In the case of civil liability rules applicable to AI-systems, aiming to improve the conditions for the

functioning of the internal market in AI-enabled products and services, the Union’s competence is

shared – according to Article 4(2)(a) TFEU, internal market as a policy area is subject to a shared

competence between the Union and its Member States.

Subsidiarity does not apply for policy areas where the Union has

exclusive

competence as defined in

Article 3 TFEU

. It is the specific legal basis which determines whether the proposal falls under the

subsidiarity control mechanism. Article 4 TFEU

sets out the areas where competence is shared

between the Union and the Member States. Article 6 TFEU

sets out the areas for which the Unions

has competence only to support the actions of the Member States.

2. Subsidiarity Principle: Why should the EU act?

2.1 Does the proposal fulfil the procedural requirements of Protocol No. 2

- Has there been a wide consultation before proposing the act?

- Is there a detailed statement with qualitative and, where possible, quantitative indicators

allowing an appraisal of whether the action can best be achieved at Union level?

An extensive consultation strategy was implemented to ensure a wide participation of stakeholders

throughout the policy cycle of this proposal. The consultation strategy was based on both a public

and several targeted consultations.

An online public consultation was open from 18 October 2021 to 10 January 2022 to gather views

from a wide variety of stakeholders, including consumers, civil society organisations, industry

associations, businesses, including SMEs, and public authorities. After analysing all the responses

Resolution of 16 February 2017 on Civil Law Rules on Robotics (2015/2103(INL)); legislative own-initiative

Resolution of 20 October 2020 on a civil liability regime for AI (2020/2014(INL)).

https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:12008E003&from=EN

https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:12008E004&from=EN

https://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=CELEX:12008E006:EN:HTML

https://eur-lex.europa.eu/legal-content/EN/TXT/HTML/?uri=CELEX:12016E/PRO/02&from=EN

kom (2022) 0496 - Ingen titel

received, the Commission published a summary outcome and the individual responses on its

website.

In addition, the proposal builds on 4 years of analysis and close involvement of stakeholders,

including academics, businesses, consumer associations, Member States and citizens. The

preparatory work started in 2018 with the setting up of the Expert Group on Liability and New

Technologies (New Technologies Formation). The Expert Group produced a Report in November

2019

that assessed the challenges some characteristics of AI pose to national civil liability rules.

The input from the Expert Group report was complemented by three additional external studies: a

comparative law study

, a behavioural economics study

and an economic study

The explanatory memorandum (section 2) and the impact assessment (chapter 3) explain how the

principle of subsidiarity has been taken into account by this legislative proposal – for further details

see answer to question 2.2 below.

2.2 Does the explanatory memorandum (and any impact assessment) accompanying the

Commission’s proposal contain an adequate justification regarding the conformity with the

principle of subsidiarity?

The explanatory memorandum clarifies that the objectives of this proposal cannot be adequately

achieved at national level because emerging divergent national rules would increase legal

uncertainty and fragmentation, creating obstacles to the rollout of AI-enabled products and services

across the internal market. Legal uncertainty would particularly affect companies active cross-border

by imposing the need for additional legal information/representation, risk management costs and

foregone revenue. At the same time, differing national rules on compensation claims for damage

caused by AI would increase transaction costs for businesses, especially for cross-border trade,

entailing significant internal market barriers. Further, legal uncertainty and fragmentation

disproportionately affect start-ups and SMEs, which account for most companies and the major

share of investments in the relevant markets.

In the absence of EU harmonised rules for compensating damage caused by AI systems, providers,

operators and users of AI systems on the one hand and injured persons on the other hand would be

faced with 27 different liability regimes, leading to different levels of protection and distorted

competition among businesses from different Member States.

Harmonised measures at EU level would significantly improve conditions for the rollout and

development of AI-technologies in the internal market by preventing fragmentation and increasing

legal certainty. This added value would be generated notably through reduced fragmentation and

increased legal certainty regarding stakeholders’ liability exposure. Moreover, only EU action can

consistently achieve the desired effect of promoting consumer trust in AI-enabled products and

services by preventing liability gaps linked to the specific characteristics of AI across the internal

market. This would ensure a consistent (minimum) level of protection for all victims (individuals and

companies) and consistent incentives to prevent damage and ensure accountability.

https://ec.europa.eu/info/law/better-regulation/have-your-say/initiatives/12979-Civil-liability-adapting-

liability-rules-to-the-digital-age-and-artificial-intelligence/public-consultation_en

Liability for artificial intelligence and other emerging digital technologies, November 2019,

https://op.europa.eu/en/publication-detail/-/publication/1c5e30be-1197-11ea-8c1f-01aa75ed71a1/language-

Karner/Koch/Geistfeld, Comparative Law Study on Civil Liability for Artificial Intelligence, 2021,

https://op.europa.eu/en/publication-detail/-/publication/8a32ccc3-0f83-11ec-9151-01aa75ed71a1/language-

en.

Kantar, Behavioural Study on the link between challenges of Artificial Intelligence for Member States’ civil

liability rules and consumer attitudes towards AI-enabled products and services, Final Report 2021.

Deloitte, Study to support the Commission’s Impact Assessment on liability for artificial intelligence, 2021.

kom (2022) 0496 - Ingen titel

The arguments above are substantiated in more detail in chapter 3 of the impact assessment.

2.3 Based on the answers to the questions below, can the objectives of the proposed action be

achieved sufficiently by the Member States acting alone (necessity for EU action)?

Given the existing differences between tort liability regimes across Member States and the likelihood

that those differences would only increase if measures were adopted at national level to adapt such

liability regimes to AI risks, the objectives of the proposed action (such as ensuring legal certainty on

a cross-border basis and consistently preventing compensation gaps in cases where AI systems are

involved) cannot be sufficiently achieved by Member States acting alone. In absence of EU action, the

identified obstacles to a well-functioning internal market for safe AI-enabled products and services

would still persist.

(a) Are there significant/appreciable transnational/cross-border aspects to the problems being

tackled? Have these been quantified?

In the area covered by this proposal, emerging divergent national rules would increase legal

uncertainty and fragmentation, creating obstacles to the rollout of AI-enabled products and services

across the internal market. Legal uncertainty would particularly affect companies active cross-

borders by imposing the need for additional legal information/representation, risk management

costs and foregone revenue. At the same time, differing national rules on compensation claims for

damage caused by AI would increase transaction costs for businesses, especially for cross-border

trade, entailing significant internal market barriers.

For the quantification of the cross-border activities in this sector, see also section 6.2.6 of the

economic study.

(b) Would national action or the absence of the EU level action conflict with core objectives of

the Treaty

or significantly damage the interests of other Member States?

In the absence of EU harmonised rules for compensating damage caused by AI systems, providers,

operators and users of AI systems on the one hand and injured persons on the other hand would be

faced with 27 different liability regimes, leading to different levels of protection and distorted

competition among businesses from different Member States, which is against the core objective of

the Treaty to establish a well-functioning internal market. At the same time, the fact that victims of

harm caused by AI systems would not benefit of the same level of protection as victims of harm

caused by other technologies is against the principle of equal access to justice, hampering the trust in

AI and ultimately the uptake of AI-enabled products/services across the EU Member States.

measures?

There is concrete evidence showing that a number of Member States may take unilateral legislative

measures to address the specific challenges posed by AI with respect to liability. For example, AI

strategies adopted in Czechia, Italy, Malta, Poland and Portugal mention initiatives to clarify liability

(see also section 3.1 of the IA). However, given the already existing large divergence between

Member States’ existing civil liability rules, it is likely that any national AI-specific measures on

liability would follow existing different national approaches and therefore increase fragmentation in

the EU. Thus relying on separate measures at the level of each Member State cannot address the

main concerns identified by the present initiative.

(d) How does the problem and its causes (e.g. negative externalities, spill-over effects) vary

https://europa.eu/european-union/about-eu/eu-in-brief_en

kom (2022) 0496 - Ingen titel

across the national, regional and local levels of the EU?

Some Member States envisage already the adoption of measures on liability for AI-enabled products

and services, while other Member States have not put forward such initiatives. This divergence of

approaches will increase the already existing fragmentation of tort liability regimes among Member

States. As a consequence, some victims would be left to bear the burden of damage caused by AI,

affecting victims disproportionately and allowing for an externalisation of costs linked to the roll-out

of AI. At the same time, companies active cross-border would face a multiplicity of liability regimes,

hampering their development and innovation potential. These elements would undermine the

functioning of the internal market for safe AI-enabled products and services. This matter is less

relevant for local and regional levels within Member States, given that tort liability is in general

regulated at national level in Europe.

(e) Is the problem widespread across the EU or limited to a few Member States?

So far, businesses that want to produce, disseminate and operate AI-enabled products and services

across borders are uncertain whether and how existing liability regimes in Member States would

apply to damage caused by AI. This uncertainty concerns not only their own Member State but

particularly other Member States where they will export to or operate their products and services.

Some Member States may take unilateral legislative measures to address the specific challenges

posed by AI with respect to liability (see Member States mentioned under 2.3(c) above). It is very

likely that any national AI-specific measures on liability would follow existing different national tort

law approaches, thus increasing the barriers to the roll-out of AI-enabled products and services

across the internal market in general, and not only on some national markets.

(f) Are Member States overstretched in achieving the objectives of the planned measure?

No, the measure is proposed to support Member States in ensuring legal certainty and consistently

preventing compensation gaps in cases where AI systems are involved.

(g) How do the views/preferred courses of action of national, regional and local authorities

differ across the EU?

Tort law traditions differ significantly across Member States and their adaptation to deal with AI-

liability cases is likely to increase such differences. The latter concern for instance the standard of

proof for causation, procedural alleviations of the burden of proof, administrative law measures

assisting those charged with the burden of proof, the basis for liability, different approaches to causal

uncertainty etc. In addition, while fault-based liabilities are the liability regimes by default in Europe

and are framed very differently across Member States, some Member States also have strict liability

regimes in place.

2.4 Based on the answer to the questions below, can the objectives of the proposed action be

better achieved at Union level by reason of scale or effects of that action (EU added value)?

The objectives of the proposed action, to ensure legal certainty and consistently prevent

compensation gaps in cases where AI systems are involved, thus creating the conditions for the

deployment of AI-technologies in the internal market, can be better achieved at Union level, by

ensuring a minimum level of harmonisation. This will increase public trust and the uptake of AI-

enabled products and services, enabling companies (in particular SMEs, which are among the most

active in the AI sector) to expand their activities across borders. On the contrary, relying only on

possible measures at national level would increase legal fragmentation across the EU and will create

obstacles to cross-border activities (as explained under question 2.3 above).

kom (2022) 0496 - Ingen titel

(a) Are there clear benefits from EU level action?

The conditions for the roll-out and development of AI-technologies in the internal market can be

significantly improved by preventing fragmentation and increasing legal certainty through

harmonised measures at EU level, compared to possible adaptations of liability rules at national

level. The economic study underpinning the Impact Assessment of this proposal concluded – as a

conservative estimate – that targeted harmonisation measures on civil liability for AI would have a

positive impact of 5 to 7 % on the production value of relevant cross-border trade. This added value

would be generated notably through reduced fragmentation and increased legal certainty regarding

stakeholders’ liability exposure. This would lower stakeholders’ legal information/representation,

internal risk management and compliance costs, facilitate financial planning as well as risk estimates

for insurance purposes, and enable companies – in particular SMEs – to explore new markets across

borders.

(b) Are there economies of scale? Can the objectives be met more efficiently at EU level (larger

benefits per unit cost)? Will the functioning of the internal market be improved?

It is only through EU action that the desired effect of promoting consumer trust in AI-enabled

products and services by preventing liability gaps linked to the specific characteristics of AI can be

achieved consistently across the internal market. Harmonised adaptations of existing liability rules

are needed, ensuring a consistent (minimum) level of protection for all victims (citizens and

companies) and consistent incentives to prevent harm and ensure accountability. The Directive will

create the conditions for businesses to use cross-border opportunities and expand their AI-related

activities in the internal market. Based on the overall value of the EU AI market affected by the

liability-related problems addressed by this Directive, it is estimated that the latter will generate an

additional market value between ca. EUR 500mln and ca. EUR 1.1bln (see page 49 of the Impact

Assessment).

homogenous policy approach?

Harmonising at EU level the relevant liability rules to respond to the risks posed by AI systems would

significantly improve the conditions for the roll-out and development of AI-technologies in the

internal market. Such more homogenous approach is needed to prevent legal fragmentation (which

may result from possible adaptations of liability rules at national level) and increase legal certainty

for consumers and businesses, while preventing liability gaps and ensuring that victims of harm

caused by AI systems are duly compensated, including in cases with cross-border dimensions.

(d) Do the benefits of EU-level action outweigh the loss of competence of the Member States

and the local and regional authorities (beyond the costs and benefits of acting at national,

regional and local levels)?

The Directive follows a minimum harmonisation approach, focusing on disclosure of information and

alleviation of the burden of proof through establishment of causality presumption subject to strict

conditions. This approach, based on a minimum intervention at EU level, allows victims of damage

caused by AI systems to invoke more favourable rules of national law. Thus, national laws could

maintain, for example, reversals of the burden of proof under fault-based regimes, or national no-

fault liability regimes of which there are already a large variety in national laws, possibly applying to

damage caused by AI systems. The benefits of EU action, described under points 2.4(a)-(c) above,

outweigh the minimum loss of competence of Member States in this area.

(e) Will there be improved legal clarity for those having to implement the legislation?

kom (2022) 0496 - Ingen titel

The risks stemming from AI systems pose challenges for the current tort liability regimes across

Member States. To tackle those risks, the Directive will bring legal clarity on the right to access to

information and the correspondent disclosure obligations, as well as alleviate the burden of proof of

claimants by establishing rebuttable presumptions of causality between non-compliance with legal

duties of care and the output of the AI system that caused the damage. These rules will significantly

increase legal certainty for claimants and defendants, for legal practitioners and national courts, as

well as for relevant public authorities. At the same time, consumers and businesses will benefit of

this enhanced legal clarity, which will increase their trust in the internal market for AI-enabled

products and services. The assessment of relevant liability risks will be facilitated, as well as their

insurability.

3. Proportionality: How the EU should act

3.1 Does the explanatory memorandum (and any impact assessment) accompanying the

Commission’s proposal contain an adequate justification regarding the proportionality of the

proposal and a statement allowing appraisal of the compliance of the proposal with the

principle of proportionality?

The explanatory memorandum clarifies that the proposal is based on a staged approach. In the first

stage, the objectives are achieved with a minimally invasive approach; the second stage involves re-

assessing the need for more stringent or more extensive measures.

The first stage is limited to the burden-of-proof measures strictly necessary to address the AI-specific

problems identified. It builds on the substantive conditions of liability currently existing in national

rules, such as fault or causality, but focuses on targeted proof-related measures, ensuring that

victims have the same level of protection as in cases not involving AI systems. Moreover, from the

various tools available in national law for easing the burden of proof, this proposal has chosen to use

rebuttable presumptions as the least interventionist tool. Such presumptions are commonly found in

national liability systems, and they balance the interests of claimants and defendants. At the same

time they are designed to incentivise compliance with existing duties of care. However, the proposal

does not lead to a reversal of the burden of proof, to avoid exposing providers, operators and users

of AI systems to higher liability risks, which may hamper innovation and reduce the uptake of AI-

enabled products and services.

The second stage included in the proposal ensures that future technological, regulatory and

jurisprudential developments will be taken into account when re-assessing the need to harmonise

other elements of the claims for compensation, including situations where no-fault liability would be

more appropriate, as requested by the European Parliament. Such assessment would also likely

consider whether such a harmonisation would need to be coupled with mandatory insurance to

ensure effectiveness.

The elements above are substantiated in more detail in the impact assessment (see especially section

4 of annex 10 to the IA, on the comparison of the policy options in terms of proportionality).

3.2 Based on the answers to the questions below and information available from any impact

assessment, the explanatory memorandum or other sources, is the proposed action an

appropriate way to achieve the intended objectives?

Yes, this Directive, focusing on the alleviation of the burden of proof, reflects the Commission’s

targeted approach of ensuring that victims are not less, but also not more protected due to the

involvement of AI. Any shifts in the risk and cost distribution between affected stakeholders that

would go beyond counter-balancing the specific proof-related challenges of AI are thus avoided, and

Member States’ well-established liability systems are respected to the maximum extent possible. In a

second stage, under the review clause, it will be assessed whether a higher level of regulatory

intervention is needed (such as non-fault liability or mandatory insurance), subject also to the

kom (2022) 0496 - Ingen titel

evolution and deployment of AI-technologies on the market and the emergence of possible incidents

linked to AI systems.

(a) Is the initiative limited to those aspects that Member States cannot achieve satisfactorily on

their own, and where the Union can do better?

The Directive is limited to the measures strictly necessary to address the AI-specific problems

identified. In particular, it would not touch upon the substantive conditions of liability like fault or

causality (which remain in the remit of national law), but focus on targeted proof-related measures

(disclosure of information and alleviation of the burden of proof through rebuttable presumptions)

ensuring that victims have the same level of protection as in cases not involving AI. Under the review

clause, the Directive provides for re-assessing the need for more stringent or more extensive

measures at a later stage, without committing to specific outcomes of such review.

(b) Is the form of Union action (choice of instrument) justified, as simple as possible, and

coherent with the satisfactory achievement of, and ensuring compliance with the objectives

pursued (e.g. choice between regulation, (framework) directive, recommendation, or

alternative regulatory methods such as co-legislation, etc.)?

A mandatory instrument is able to prevent protection gaps stemming from partial or no

implementation. While a non-binding instrument presents a less intrusive approach, it is unlikely to

address the identified problems in an effective manner. Among the binding instruments, a directive is

the most suitable instrument, as it provides the desired harmonisation effect and legal certainty,

while also being the adequate instrument to enable MS in a flexible manner to fit the harmonised

measures without friction into their national liability regimes. On the contrary, a regulation would

not have allowed for such flexibility in the politically sensitive field of civil law. On these grounds, the

choice has been made to submit this proposal in the form of a directive.

satisfactorily the objectives set? (e.g. is it possible to limit the European action to minimum

standards or use a less stringent policy instrument or approach?)

The Directive achieves the set objectives with a minimally invasive approach. It is focused on burden-

of-proof measures strictly necessary to address the AI-specific problems identified. It builds on the

substantive conditions of liability currently existing in national rules, such as fault or causality, but

focuses on targeted proof-related measures, ensuring that victims have the same level of protection

as in cases not involving AI systems. Moreover, from the various tools available in national law for

easing the burden of proof, this proposal has chosen to use rebuttable presumptions as the least

interventionist tool. However, the proposal does not lead to a reversal of the burden of proof, to

avoid exposing providers, operators and users of AI systems to higher liability risks, which may

hamper innovation and reduce the uptake of AI-enabled products and services. Given that this

Directive follows a minimum harmonisation approach, Member States remain free to use in their

national law more invasive tools if they deem it appropriate, such as reversals of the burden of proof,

irrebuttable presumptions, strict liability etc.

(d) Does the initiative create financial or administrative cost for the Union, national

governments, regional or local authorities, economic operators or citizens? Are these costs

commensurate with the objective to be achieved?

The Directive will not have implications for the budget of the European Union or its Member States,

at national, regional or local level. It also does not impose specific costs on economic operators or

citizens. By reducing fragmentation and increasing legal certainty regarding stakeholders’ liability

exposure, the Directive lowers stakeholders’ legal information/representation, internal risk

kom (2022) 0496 - Ingen titel

management and compliance costs, facilitating financial planning as well as risk estimates for

insurance purposes.

(e) While respecting the Union law, have special circumstances applying in individual Member

States been taken into account?

By proposing a minimum harmonisation approach, this Directive takes into account the different

levels of regulatory intervention of Member States as regards the tort liability for damages caused by

AI systems. The Impact Assessment and the Comparative Legal Study underpinning this proposal

have assessed the various civil law traditions on tort liability present in different Member States. The

directive as a legal instrument gives the Member States the needed flexibility when integrating the

harmonised measures into their national liability regimes.