Télécharger la pièce jointe originale
(PDF file)

Ceci est une version HTML d'une pièce jointe de la demande d'accès à l'information 'Artificial intelligence: questions of interpretation and application of international law - EU in the areas of civil and military uses and of state authority outside the scope of criminal justice'.


European Parliament
2019-2024
Subcommittee on Security and Defence
2020/2013(INI)(INI)
30.6.2020
COMPROMISE AMENDMENT
1 - 11
Draft opinion
Urmas Paet
(PE650.702v01-00)
on artificial intelligence: questions of interpretation and application of
international law insofar as the EU is affected in the areas of civil and military
uses and of state authority outside the scope of criminal justice
(2020/2013(INI)(INI))
AM\2222222EN.docx
PE111.111v01-00
EN
United in diversity
EN
AM_Com_NonLegCompr
PE111.111v01-00
2/10
AM\2222222EN.docx
EN
Amendment 1
Urmas Paet
Compromise amendment replacing Amendment(s): 2, 3, 4, 5 and 6
Draft opinion
Paragraph 1
Draft opinion
Amendment
1.
Highlights that the security and
1.
Highlights that the security and
defence policies of the European Union
defence policies of the European Union
and its Member States are guided by the
and its Member States are guided by the
principles of the UN Charter, and by a
principles enshrined in the European
common understanding of the universal
Charter of Fundamental Rights and of
values of the inviolable and inalienable
those of the UN Charter, which calls upon
rights of the human person, of freedom, of
all States to refrain from the threat or use
democracy, of equality and of the rule of
of force in their relation with each other,
law; highlights that all defence-related
as well as by the international law, by the
efforts within the Union framework must
principles of human rights and respect for
respect these universal values while
human dignity and by a common
promoting peace, security and progress in
understanding of the universal values of
Europe and in the world;
the inviolable and inalienable rights of the
human person, of freedom, of democracy,
of equality and of the rule of law;
highlights that all defence-related efforts
within the Union framework must respect
these universal values while promoting
peace, stability, security and progress in
Europe and in the world;
Or. en
Amendment 2
Urmas Paet
Compromise amendment replacing Amendment(s): 7, 8, 9, 10, 17, 20, 66 and 67
Draft opinion
Paragraph 2
Draft opinion
Amendment
2.
Calls on the UN and the wider
2.
Calls on the UN and the wider
international community to undertake all
international community to undertake all
necessary efforts to ensure that the
necessary regulatory efforts to ensure that
application of Artificial Intelligence (AI) in
the development and application of
military affairs and the use of AI-enabled
Artificial Intelligence (AI) in military
systems by the military stay within the
affairs andin the law enforcement civil
AM\2222222EN.docx
3/10
PE111.111v01-00
EN
strict limits set by international law and
capacities, such as police and border
international humanitarian law (IHL);
control forces and the study, development
and use of AI-enabled systems by the
military stay within the strict limits set by
international law, including international
humanitarian law (IHL) and Human
Rights Law; stresses that the EU should
pursue the international adoption of its
technical and ethical standards in AI-
enabled military systems, and, in close
cooperation with like-minded partners,
strive for an international regulatory
framework, agreeing on common norms,
based on democratic values, adequately
framed as to prevent their use for
espionage, mass, targeted and political
surveillance, disinformation and data
manipulation, and a cyber arms race;
calls for an increased cooperation with
NATO Alliance for the establishment of
common standards and interoperability of
AI-enabled systems; calls on the
European Commission to foster dialogue,
closer cooperation and synergies among
Member States, researchers, academics,
civil society actors and the private sector,
in particular leading companies and
enterprises, and the military so as to have
inclusive policymaking processes when it
comes to defence-related AI regulations;
Or. en
Amendment 3
Urmas Paet
Compromise amendment replacing Amendment(s): 14, 15 and 16
Draft opinion
Paragraph 3
Draft opinion
Amendment
3.
Considers in particular that AI-
3.
Considers in particular that the
enabled systems must abide by the
design, development and the use of AI-
principles of the Martens Clause, and must
enabled systems in armed conflicts must,
never breach or be permitted to breach the
as provided by the Martens Clause, abide
dictates of the public conscience and
by the general principles of IHL and must
humanity; considers that this is the
never breach or be permitted to breach the
PE111.111v01-00
4/10
AM\2222222EN.docx
EN
ultimate test for the admissibility of an AI-
dictates of the public conscience and
enabled system in warfare; calls on the AI
humanity; considers that their ability to be
research community to integrate this
used in compliance with international
principle in all AI-enabled systems
humanitarian law is the minimum
intended to be used in warfare; considers
standard for the admissibility of an AI-
that no authority can issue a derogation
enabled system in warfare; calls on the AI
from those principles or certify an AI-
research community to integrate this
enabled system;
principle in all AI-enabled systems
intended to be used in warfare; considers
that no authority can issue a derogation
from those principles or certify an AI-
enabled system breaching them;
Or. en
Amendment 4
Urmas Paet
Compromise amendment replacing Amendment(s): 21, 22, 23, 24, 25 and 27
Draft opinion
Paragraph 4
Draft opinion
Amendment
4.
Stresses that states, parties to a
4.
Stresses that states, parties to a
conflict and individuals, when employing
conflict and individuals, when employing
AI-enabled systems in warfare, must at all
AI-enabled systems in warfare, must at all
times adhere to their obligations under the
times adhere to their obligations and
applicable international law and remain
liability under the applicable international
accountable for actions resulting from the
law and must remain accountable for
use of such systems; recalls that AI
actions resulting from the use of such
machines can under no circumstances be
systems; recalls that humans remain
held accountable for intended, unintended
accountable for intended, unintended or
or undesirable effects caused by AI-
undesirable effects caused by AI-enabled
enabled systems on the battlefield;
systems on the battlefield; emphasizes that
the decision to take lethal action by means
of weapons systems with a high degree of
autonomy must always be made by human
operators exercising meaningful control
and oversight and necessary level of
judgment in line with the principles of
proportionality and necessity; stresses and
that AI-enabled systems can under no
circumstances be permitted to replace
human decision;
Or. en
AM\2222222EN.docx
5/10
PE111.111v01-00
EN
Amendment 5
Urmas Paet
Compromise amendment replacing Amendment(s): 28, 29, 30 and 33
Draft opinion
Paragraph 5
Draft opinion
Amendment
5.
Highlights the need to take duly
5.
Highlights the need to take duly
into account, during the design,
into account, during the design,
development, testing and deployment
development, testing and deployment
phases of an AI-enabled system, potential
phases of an AI-enabled system, potential
risks as regards civilian casualties and
risks as regards, in particular to incidental
injury, accidental loss of life, and damage
civilian casualties and injury, accidental
to civilian infrastructure, but also risks
loss of life, and damage to civilian
related to unintended engagement,
infrastructure, but also risks related to
manipulation, proliferation, cyber-attack or
unintended engagement,
interference and acquisition by terrorist
manipulation,proliferation, cyber-attack or
groups;
interference and acquisition by organised
crime and terrorist groups, lading to
escalatory destabilising effects that aim to
make our societies, militaries and
institutions vulnerable;
Or. en
Amendment 6
Urmas Paet
Compromise amendment replacing Amendment(s): 30 and 33
Draft opinion
Paragraph 5 a (new)
Draft opinion
Amendment
5 a.
Is concerned by the fact that not
all Members of the international
community would follow a regulatory
human-centric AI approach; urges the
EU and Member States to assess the
development of AI technologies,
particularly military and surveillance,
within authoritarian states that avoid
compliance with EU led regulations;
Or. en
PE111.111v01-00
6/10
AM\2222222EN.docx
EN
Amendment 7
Urmas Paet
Compromise amendment replacing Amendment(s): 31, 35, 36 and 37
Draft opinion
Paragraph 6
Draft opinion
Amendment
6.
Stresses the need for robust testing
6.
Stresses the need for robust testing,
and evaluation systems based on norms to
evaluation, certification, monitoring and
ensure that during the entire lifecycle of
verification systems based on clear legal
AI-enabled systems in the military domain,
and democratic norms as well as on clear
in particular during the phases of human-
safety and security provisions, to ensure
machine interaction, machine learning and
that during the entire life cycle of AI-
adjusting and adapting to new
enabled systems in the military domain, in
circumstances, the systems do not go
particular during the phases of human-
beyond the intended limits and will at all
machine interaction, machine learning and
times comply with the applicable
adjusting and adapting to new
international law;
circumstances, the systems and their effect
do not go beyond the intended limits and
must be used at all times incompliance
with the applicable international law;
Or. en
Amendment 8
Urmas Paet
Compromise amendment replacing Amendment(s): 40, 41, 43 and 44
Draft opinion
Paragraph 7
Draft opinion
Amendment
7.
Highlights that any AI-enabled
7.
Highlights that any AI-enabled
system used in the military domain must,
system used in the military domain must,
as a minimum set of requirements, be able
as a minimum set of requirements, be able
to distinguish between combatants and
to distinguish between combatants and
non-combatants on the battlefield, not have
non-combatants on the battlefield, between
indiscriminate effects, not cause
military and civilian targets, recognize
unnecessary suffering to persons, not be
when a combatant surrenders or is hors
biased or be trained on biased data, and be
de combat, not have indiscriminate effects,
in compliance with the IHL principles of
individuate the use of force and not target
military necessity, proportionality in the
a certain category of people, not be of a
AM\2222222EN.docx
7/10
PE111.111v01-00
EN
use of force and precaution prior to
nature to cause unnecessary suffering to
engagement;
persons, nor cruel or degrading
treatments, not be biased or be trained on
biased data, and be used in compliance
with the IHL general principles of
humanity, distinction, proportionality,
precaution and the principle of military
necessity prior to engagement and in
attack; underlines the importance of the
quality of algorithms, original data and
ex-ante review of decision-making
processes;
Or. en
Amendment 9
Urmas Paet
Compromise amendment replacing Amendment(s): 46, 47 and 48
Draft opinion
Paragraph 8
Draft opinion
Amendment
8.
Stresses that in the use of AI-
8.
Stresses that in the use of AI-
enabled systems in security and defence,
enabled systems in security and defence,
full situational understanding of the
comprehensive situational understanding
operator, ability to detect possible changes
of the human operator, predictability and
in circumstances and ability to discontinue
reliability of the AI-enabled system, as
an attack are needed to ensure that IHL
well as the human operator’s ability to
principles, in particular distinction,
detect possible changes in circumstances
proportionality and precaution in attack,
and operational environment and ability
are fully applied across the entire chain of
to intervene in or discontinue an attack are
command and control; stresses that AI-
needed to ensure that IHL principles, in
enabled systems must allow the military
particular distinction, proportionality and
leadership to assume its full responsibility
precaution in attack, are fully applied
at all times;
across the entire chain of command and
control; stresses that AI-enabled systems
must allow the military leadership to exert
meaningful control, to assume its full
responsibility and be accountable
throughout each of their uses;
Or. en
Amendment 10
Urmas Paet
PE111.111v01-00
8/10
AM\2222222EN.docx
EN
Compromise amendment replacing Amendment(s): 53 and 54
Draft opinion
Paragraph 9
Draft opinion
Amendment
9.
Calls on states to carry out an
9.
Encourages states to carry out an
assessment of how autonomous military
assessment on whether and how
devices have contributed to their national
autonomous military device shave
security and what their national security
contributed to their national security and
can gain from AI-enabled weapon
what their national security could gain
systems, in particular as regards the
from AI-enabled weapon systems, in
potential of such technologies to reduce
particular as regards the potential of such
human error, thus enhancing the
technologies to support and enhance
implementation of IHL principles;
human decision-making in compliance
with IHL and its principles;
Or. en
Amendment 11
Urmas Paet
Compromise amendment replacing Amendment(s): 20, 57, 58, 59 and 62
Draft opinion
Paragraph 10
Draft opinion
Amendment
10.
Calls on the HR/VP, in the
10.
Recalls the European Parliament’s
framework of the ongoing discussions on
position on autonomous weapons systems
the international regulation of lethal
to not enable strikes to be carried out
autonomous weapon systems by states
without meaningful human intervention
parties to the Convention on Prohibitions
of 12 September 2018; calls on the
or Restrictions on the Use of Certain
HR/VP, the Member States and the
Conventional Weapons (CCW), to help
European Council to adopt a common
streamline the global debate on core
position on autonomous weapons system,
issues and definitions where consensus
that ensures meaningful human control
has not been reached, in particular as
over the critical functions of weapons
regards concepts and characteristics of AI-
systems, including during deployment;
enabled lethal autonomous weapons and
reaffirms support to the work of the UN
their functions in the identification,
CCWGGE on LAWs, which remains the
selection and engagement of a target,
relevant international forum for
application of the concept of human
discussions and negotiations on the legal
responsibility in the use of AI-enabled
challenges posed by autonomous weapons
systems in defence, and the degree of
systems; calls for all existing multilateral
human/machine interaction, including the
efforts to be accelerated so that normative
AM\2222222EN.docx
9/10
PE111.111v01-00
EN
concept of human control and judgment,
and regulatory frameworks are not
during the different stages of the lifecycle
outpaced by technological development
of an AI-enabled weapon.
and new methods of warfare; calls on the
HR/VP, in the framework of the ongoing
discussions on the international regulation
of lethal autonomous weapon systems by
states parties to the Convention on
Prohibitions or Restrictions on the Use of
Certain Conventional Weapons (CCW) to
remain engaged and help
advancing,without delay, the effort to
develop a new global normative
framework and a legally binding
instrument, focused on definitions,
concepts and characteristics of emerging
technologies in the area of lethal
autonomous weapons systems, ethical and
legal questions of human control, in
particular with regard to their critical
functions, such as target selection and
engagement, retention of human
responsibility and accountability and the
necessary degree of human/machine
interaction,including the concept of human
control and human judgment, to ensure
compliance with international
humanitarian and Human Rights law
during the different stages of the lifecycle
of an AI-enabled weapon, with a view to
agree tangible recommendations on the
clarification, consideration and
development of aspects of the normative
framework on emerging technologies in
the area of lethal autonomous weapons
systems;
Or. en
PE111.111v01-00
10/10
AM\2222222EN.docx
EN