What changed between the plenary report and the adopted text

9 unchanged paragraphs

Recital 18: (18) The use of AI systems for ‘real-time’ remote biometric identification of natural persons in publicly accessible spaces is particularly intrusive to the rights and freedoms of the concerned persons, and can ultimately affect the private life of a large part of the population, evoke a feeling of constant surveillance, give parties deploying biometric identification in publicly accessible spaces a position of uncontrollable power and indirectly dissuade the exercise of the freedom of assembly and other fundamental rights at the core to the Rule of Law. Technical inaccuracies of AI systems intended for the remote biometric identification of natural persons can lead to biased results and entail discriminatory effects. This is particularly relevant when it comes to age, ethnicity, sex or disabilities. In addition, the immediacy of the impact and the limited opportunities for further checks or corrections in relation to the use of such systems operating in ‘real-time’ carry heightened risks for the rights and freedoms of the persons that are concerned by law enforcement activities. The use of those systems in publicly accessible places should therefore be prohibited. Similarly, AI systems used for the analysis of recorded footage of publicly accessible spaces through ‘post’ remote biometric identification systems should also be prohibited, unless there is pre-judicial authorisation for use in the context of law enforcement, when strictly necessary for the targeted search connec…

Recital 19: deleted / (deleted)

Recital 20: deleted

Recital 21: deleted

Recital 22: deleted

Recital 23: deleted

Recital 24: (24) Any processing of biometric data and other personal data involved in the use of AI systems for biometric identification, other than in connection to the use of ‘real-time’ remote biometric identification systems in publicly accessible spaces as regulated by this Regulation should continue to comply with all requirements resulting from Article 9(1) of Regulation (EU) 2016/679, Article 10(1) of Regulation (EU) 2018/1725 and Article 10 of Directive (EU) 2016/680, as applicable.

Recital 25: (25) In accordance with Article 6a of Protocol No 21 on the position of the United Kingdom and Ireland in respect of the area of freedom, security and justice, as annexed to the TEU and to the TFEU, Ireland is not bound by the rules laid down in Article 5(1), point (d), of this Regulation adopted on the basis of Article 16 of the TFEU which relate to the processing of personal data by the Member States when carrying out activities falling within the scope of Chapter 4 or Chapter 5 of Title V of Part Three of the TFEU, where Ireland is not bound by the rules governing the forms of judicial cooperation in criminal matters or police cooperation which require compliance with the provisions laid down on the basis of Article 16 of the TFEU.

Recital 26: (26) In accordance with Articles 2 and 2a of Protocol No 22 on the position of Denmark, annexed to the TEU and TFEU, Denmark is not bound by rules laid down in Article 5(1), point (d) of this Regulation adopted on the basis of Article 16 of the TFEU, or subject to their application, which relate to the processing of personal data by the Member States when carrying out activities falling within the scope of Chapter 4 or Chapter 5 of Title V of Part Three of the TFEU.

Change 23

Changed:Recital 26 a (new): (26 a)(26a) AI systems used by law enforcement authorities or on their behalf to make predictions, profiles or risk assessments based on profiling of natural persons or data analysis based on personality traits and characteristics, including the person’s location, or past criminal behaviour of natural persons or groups of persons for the purpose of predicting the occurrence or reoccurrence of an actual or potential criminal offence(s) or other criminalised social behaviour or administrative offences, including fraud-predicition systems, hold a particular risk of discrimination against certain persons or groups of persons, as they violate human dignity as well as the key legal principle of presumption of innocence. Such AI systems should therefore be prohibited.

Change 24

Changed:Recital 26 b (new): (26 b)(26b) The indiscriminate and untargeted scraping of biometric data from social media or CCTV footage to create or expand facial recognition databases add to the feeling of mass surveillance and can lead to gross violations of fundamental rights, including the right to privacy. The use of AI systems with this intended purpose should therefore be prohibited.

Change 25

Changed:Recital 26 c (new): (26 c)(26c) There are serious concerns about the scientific basis of AI systems aiming to detect emotions, physical or physiological features such as facial expressions, movements, pulse frequency or voice. Emotions or expressions of emotions and perceptions thereof vary considerably across cultures and situations, and even within a single individual. Among the key shortcomings of such technologies, are the limited reliability (emotion categories are neither reliably expressed through, nor unequivocally associated with, a common set of physical or physiological movements), the lack of specificity (physical or physiological expressions do not perfectly match emotion categories) and the limited generalisability (the effects of context and culture are not sufficiently considered). Reliability issues and consequently, major risks for abuse, may especially arise when deploying the system in real-life situations related to law enforcement, border management, workplace and education institutions. Therefore, the placing on the market, putting into service, or use of AI systems intended to be used in these contexts to detect the emotional state of individuals should be prohibited.

Change 26

Changed:Recital 26 d (new): (26 d)(26d) Practices that are prohibited by Union legislation, including data protection law, non-discrimination law, consumer protection law, and competition law, should not be affected by this Regulation

Recital 27: (27) High-risk AI systems should only be placed on the Union market, put into service or used if they comply with certain mandatory requirements. Those requirements should ensure that high-risk AI systems available in the Union or whose output is otherwise used in the Union do not pose unacceptable risks to important Union public interests as recognised and protected by Union law, including fundamental rights, democracy, the rule or law or the environment. In order to ensure alignment with sectoral legislation and avoid duplications, requirements for high-risk AI systems should take into account sectoral legislation laying down requirements for high-risk AI systems included in the scope of this Regulation, such as Regulation (EU) 2017/745 on Medical Devices and Regulation (EU) 2017/746 on In Vitro Diagnostic Devices or Directive 2006/42/EC on Machinery. AI systems identified as high-risk should be limited to those that have a significant harmful impact on the health, safety and fundamental rights of persons in the Union and such limitation minimises any potential restriction to international trade, if any. Given the rapid pace of technological development, as well as the potential changes in the use of AI systems, the list of high-risk areas and use-cases in Annex III should nonetheless be subject to permanent review through the exercise of regular assessment.

Recital 28: (28) AI systems could have an adverse impact to health and safety of persons, in particular when such systems operate as safety components of products. Consistently with the objectives of Union harmonisation legislation to facilitate the free movement of products in the internal market and to ensure that only safe and otherwise compliant products find their way into the market, it is important that the safety risks that may be generated by a product as a whole due to its digital components, including AI systems, are duly prevented and mitigated. For instance, increasingly autonomous robots, whether in the context of manufacturing or personal assistance and care should be able to safely operate and performs their functions in complex environments. Similarly, in the health sector where the stakes for life and health are particularly high, increasingly sophisticated diagnostics systems and systems supporting human decisions should be reliable and accurate.

Change 27

Changed:Recital 28 a (new): (28 a)(28a) The extent of the adverse impact caused by the AI system on the fundamental rights protected by the Charter is of particular relevance when classifying an AI system as high-risk. Those rights include the right to human dignity, respect for private and family life, protection of personal data, freedom of expression and information, freedom of assembly and of association, and non-discrimination, right to education consumer protection, workers’ rights, rights of persons with disabilities, gender equality, intellectual property rights, right to an effective remedy and to a fair trial, right of defence and the presumption of innocence, right to good administration. In addition to those rights, it is important to highlight that children have specific rights as enshrined in Article 24 of the EU Charter and in the United Nations Convention on the Rights of the Child (further elaborated in the UNCRC General Comment No. 25 as regards the digital environment), both of which require consideration of the children’s vulnerabilities and provision of such protection and care as necessary for their well-being. The fundamental right to a high level of environmental protection enshrined in the Charter and implemented in Union policies should also be considered when assessing the severity of the harm that an AI system can cause, including in relation to the health and safety of persons or to the environment.

4 unchanged paragraphs

Recital 29: (29) As regards high-risk AI systems that are safety components of products or systems, or which are themselves products or systems falling within the scope of Regulation (EC) No 300/2008 of the European Parliament and of the Council39 , Regulation (EU) No 167/2013 of the European Parliament and of the Council40 , Regulation (EU) No 168/2013 of the European Parliament and of the Council41 , Directive 2014/90/EU of the European Parliament and of the Council42 , Directive (EU) 2016/797 of the European Parliament and of the Council43 , Regulation (EU) 2018/858 of the European Parliament and of the Council44 , Regulation (EU) 2018/1139 of the European Parliament and of the Council45 , and Regulation (EU) 2019/2144 of the European Parliament and of the Council46 , it is appropriate to amend those acts to ensure that the Commission takes into account, on the basis of the technical and regulatory specificities of each sector, and without interfering with existing governance, conformity assessment, market surveillance and enforcement mechanisms and authorities established therein, the mandatory requirements for high-risk AI systems laid down in this Regulation when adopting any relevant future delegated or implementing acts on the basis of those acts.

Recital 30: (30) As regards AI systems that are safety components of products, or which are themselves products, falling within the scope of certain Union harmonisation law listed in Annex II, it is appropriate to classify them as high-risk under this Regulation if the product in question undergoes the conformity assessment procedure in order to ensure compliance with essential safety requirements with a third-party conformity assessment body pursuant to that relevant Union harmonisation law. In particular, such products are machinery, toys, lifts, equipment and protective systems intended for use in potentially explosive atmospheres, radio equipment, pressure equipment, recreational craft equipment, cableway installations, appliances burning gaseous fuels, medical devices, and in vitro diagnostic medical devices.

Recital 31: (31) The classification of an AI system as high-risk pursuant to this Regulation should not mean that the product whose safety component is the AI system, or the AI system itself as a product, is considered ‘high-risk’ under the criteria established in the relevant Union harmonisation law that applies to the product. This is notably the case for Regulation (EU) 2017/745 of the European Parliament and of the Council47 and Regulation (EU) 2017/746 of the European Parliament and of the Council48 , where a third-party conformity assessment is provided for medium-risk and high-risk products.

Recital 32: (32) As regards stand-alone AI systems, meaning high-risk AI systems other than those that are safety components of products, or which are themselves products and that are listed in one of the areas and use cases in Annex III, it is appropriate to classify them as high-risk if, in the light of their intended purpose, they pose a significant risk of harm to the health and safety or the fundamental rights of persons and, where the AI system is used as a safety component of a critical infrastructure, to the environment . Such significant risk of harm should be identified by assessing on the one hand the effect of such risk with respect to its level of severity, intensity, probability of occurrence and duration combined altogether and on the other hand whether the risk can affect an individual, a plurality of persons or a particular group of persons. Such combination could for instance result in a high severity but low probability to affect a natural person, or a high probability to affect a group of persons with a low intensity over a long period of time, depending on the context. The identification of those systems is based on the same methodology and criteria envisaged also for any future amendments of the list of high-risk AI systems.

Change 28

Changed:Recital 32 a (new): (32 a)(32a) Providers whose AI systems fall under one of the areas and use cases listed in Annex III that consider their system does not pose a significant risk of harm to the health, safety, fundamental rights or the environment should inform the national supervisory authorities by submitting a reasoned notification. This could take the form of a one-page summary of the relevant information on the AI system in question, including its intended purpose and why it would not pose a significant risk of harm to the health, safety, fundamental rights or the environment. The Commission should specify criteria to enable companies to assess whether their system would pose such risks, as well as develop an easy to use and standardised template for the notification. Providers should submit the notification as early as possible and in any case prior to the placing of the AI system on the market or its putting into service, ideally at the development stage, and they should be free to place it on the market at any given time after the notification. However, if the authority estimates the AI system in question was misclassified, it should object to the notification within a period of three months. The objection should be substantiated and duly explain why the AI system has been misclassified. The provider should retain the right to appeal by providing further arguments. If after the three months there has been no objection to the notification, national supervisory authorities could still interve…interven…

Recital 33: deleted

Change 29

Changed:Recital 33 a (new): (33 a)(33a) As biometric data constitute a special category of sensitive personal data in accordance with Regulation 2016/679, it is appropriate to classify as high-risk several critical use-cases of biometric and biometrics-based systems. AI systems intended to be used for biometric identification of natural persons and AI systems intended to be used to make inferences about personal characteristics of natural persons on the basis of biometric or biometrics-based data, including emotion recognition systems, with the exception of those which are prohibited under this Regulation should therefore be classified as high-risk. This should not include AI systems intended to be used for biometric verification, which includes authentication, whose sole purpose is to confirm that a specific natural person is the person he or she claims to be and to confirm the identity of a natural person for the sole purpose of having access to a service, a device or premises (one-to-one verification). Biometric and biometrics-based systems which are provided for under Union law to enable cybersecurity and personal data protection measures should not be considered as posing a significant risk of harm to the health, safety and fundamental rights.

Change 30

Changed:Recital 34: (34) As regards the management and operation of critical infrastructure, it is appropriate to classify as high-risk the AI systems intended to be used as safety components in the management and operation of the supply of water, gas, heating electricity and critical digital infrastructure, since their failure or malfunctioning may infringe the security and integrity of such critical infrastructure or put at risk the life and health of persons at large scale and lead to appreciable disruptions in the ordinary conduct of social and economic activities. Safety components of critical infrastructure, including critical digital infrastructure, are systems used to directly protect the physical integrity of physicalcritical infrastructure or health and safety of persons and property. Failure or malfunctioning of such components might directly lead to risks to the physical integrity of critical infrastructure and thus to risks to the health and safety of persons and property. Components intended to be used solely for cybersecurity purposes should not qualify as safety components. Examples of such safety components may include systems for monitoring water pressure or fire alarm controlling systems in cloud computing centres.

Recital 35: (35) Deployment of AI systems in education is important in order to help modernise entire education systems, to increase educational quality, both offline and online and to accelerate digital education, thus also making it available to a broader audience . AI systems used in education or vocational training, notably for determining access or materially influence decisions on admission or assigning persons to educational and vocational training institutions or to evaluate persons on tests as part of or as a precondition for their education or to assess the appropriate level of education for an individual and materially influence the level of education and training that individuals will receive or be able to access or to monitor and detect prohibited behaviour of students during tests should be classified as high-risk AI systems, since they may determine the educational and professional course of a person’s life and therefore affect their ability to secure their livelihood. When improperly designed and used, such systems can be particularly intrusive and may violate the right to education and training as well as the right not to be discriminated against and perpetuate historical patterns of discrimination, for example against women, certain age groups, persons with disabilities, or persons of certain racial or ethnic origins or sexual orientation.

Change 31

Changed:Recital 36: (36) AI systems used in employment, workers management and access to self-employment, notably for the recruitment and selection of persons, for making decisions or materially influence decisions on initiation, promotion and termination and for personalised task allocation based on individual behaviour, personal traits or biometric data, monitoring or evaluation of persons in work-related contractual relationships, should also be classified as high-risk, since those systems may appreciably impact future career prospects, livelihoods of these persons and workers’ rights. Relevant work-related contractual relationships should meaningfully involve employees and persons providing services through platforms as referred to in the Commission Work Programme 2021. Throughout the recruitment process and in the evaluation, promotion, or retention of persons in work-related contractual relationships, such systems may perpetuate historical patterns of discrimination, for example against women, certain age groups, persons with disabilities, or persons of certain racial or ethnic origins or sexual orientation. AI systems used to monitor the performance and behaviour of these persons may also undermine the essence of their fundamental impact their rights to data protection and privacy. This Regulation applies without prejudice to Union and Member State competences to provide for more specific rules for the use of AI-systems in the employment context.

Recital 37: (37) Another area in which the use of AI systems deserves special consideration is the access to and enjoyment of certain essential private and public services, including healthcare services, and essential services, including but not limited to housing, electricity, heating/cooling and internet, and benefits necessary for people to fully participate in society or to improve one’s standard of living. In particular, AI systems used to evaluate the credit score or creditworthiness of natural persons should be classified as high-risk AI systems, since they determine those persons’ access to financial resources or essential services such as housing, electricity, and telecommunication services. AI systems used for this purpose may lead to discrimination of persons or groups and perpetuate historical patterns of discrimination, for example based on racial or ethnic origins, gender, disabilities, age, sexual orientation, or create new forms of discriminatory impacts. However, AI systems provided for by Union law for the purpose of detecting fraud in the offering of financial services should not be considered as high-risk under this Regulation. Natural persons applying for or receiving public assistance benefits and services from public authorities, including healthcare services and essential services, including but not limited to housing, electricity, heating/cooling and internet, are typically dependent on those benefits and services and in a vulnerable position in relation to the r…

Change 32

Changed:Recital 37 a (new): (37 a)(37a) Given the role and responsibility of police and judicial authorities, and the impact of decisions they take for the purposes of the prevention, investigation, detection or prosecution of criminal offences or the execution of criminal penalties, some specific use-cases of AI applications in law enforcement has to be classified as high-risk, in particular in instances where there is the potential to significantly affect the lives or the fundamental rights of individuals.

Recital 38: (38) Actions by law enforcement authorities involving certain uses of AI systems are characterised by a significant degree of power imbalance and may lead to surveillance, arrest or deprivation of a natural person’s liberty as well as other adverse impacts on fundamental rights guaranteed in the Charter. In particular, if the AI system is not trained with high quality data, does not meet adequate requirements in terms of its performance, its accuracy or robustness, or is not properly designed and tested before being put on the market or otherwise put into service, it may single out people in a discriminatory or otherwise incorrect or unjust manner. Furthermore, the exercise of important procedural fundamental rights, such as the right to an effective remedy and to a fair trial as well as the right of defence and the presumption of innocence, could be hampered, in particular, where such AI systems are not sufficiently transparent, explainable and documented. It is therefore appropriate to classify as high-risk a number of AI systems intended to be used in the law enforcement context where accuracy, reliability and transparency is particularly important to avoid adverse impacts, retain public trust and ensure accountability and effective redress. In view of the nature of the activities in question and the risks relating thereto, those high-risk AI systems should include in particular AI systems intended to be used by or on behalf of law enforcement authorities or by Union agenc…

Recital 39: (39) AI systems used in migration, asylum and border control management affect people who are often in particularly vulnerable position and who are dependent on the outcome of the actions of the competent public authorities. The accuracy, non-discriminatory nature and transparency of the AI systems used in those contexts are therefore particularly important to guarantee the respect of the fundamental rights of the affected persons, notably their rights to free movement, non-discrimination, protection of private life and personal data, international protection and good administration. It is therefore appropriate to classify as high-risk AI systems intended to be used by or on behalf of competent public authorities or by Union agencies, offices or bodies charged with tasks in the fields of migration, asylum and border control management as polygraphs and similar tools insofar as their use is permitted under relevant Union and national law, for assessing certain risks posed by natural persons entering the territory of a Member State or applying for visa or asylum; for verifying the authenticity of the relevant documents of natural persons; for assisting competent public authorities for the examination and assessment of the veracity of evidence in relation to applications for asylum, visa and residence permits and associated complaints with regard to the objective to establish the eligibility of the natural persons applying for a status; for monitoring, surveilling or processing …

Change 33

Changed:Recital 40: (40) Certain AI systems intended for the administration of justice and democratic processes should be classified as high-risk, considering their potentially significant impact on democracy, rule of law, individual freedoms as well as the right to an effective remedy and to a fair trial. In particular, to address the risks of potential biases, errors and opacity, it is appropriate to qualify as high-risk AI systems intended to be used by a judicial authority or administrative body or on their behalf to assist judicial authorities or administrative bodies in researching and interpreting facts and the law and in applying the law to a concrete set of facts or used in a similar way in alternative dispute resolution. The use of artificial intelligence tools can support, but should not replace the decision-making power of judges or judicial independence, as the final decision-making must remain a human-driven activity and decision. Such qualification should not extend, however, to AI systems intended for purely ancillary administrative activities that do not affect the actual administration of justice in individual cases, such as anonymisation or pseudonymisation of judicial decisions, documents or data, communication between personnel, administrative tasks or allocation of resources.

Change 34

Changed:Recital 40 a (new): (40 a)(40a) In order to address the risks of undue external interference to the right to vote enshrined in Article 39 of the Charter, and of disproportionate effects on democratic processes, democracy, and the rule of law, AI systems intended to be used to influence the outcome of an election or referendum or the voting behaviour of natural persons in the exercise of their vote in elections or referenda should be classified as high-risk AI systems. with the exception of AI systems whose output natural persons are not directly exposed to, such as tools used to organise, optimise and structure political campaigns from an administrative and logistical point of view.

Change 35

Changed:Recital 40 b (new): (40 b)(40b) Considering the scale of natural persons using the services provided by social media platforms designated as very large online platforms, such online platforms can be used in a way that strongly influences safety online, the shaping of public opinion and discourse, election and democratic processes and societal concerns. It is therefore appropriate that AI systems used by those online platforms in their recommender systems are subject to this Regulation so as to ensure that the AI systems comply with the requirements laid down under this Regulation, including the technical requirements on data governance, technical documentation and traceability, transparency, human oversight, accuracy and robustness. Compliance with this Regulation should enable such very large online platforms to comply with their broader risk assessment and risk-mitigation obligations in Article 34 and 35 of Regulation EU 2022/2065. The obligations in this Regulation are without prejudice to Regulation (EU) 2022/2065 and should complement the obligations required under the Regulation (EU) 2022/2065 when the social media platform has been designated as a very large online platform. Given the European-wide impact of social media platforms designated as very large online platforms, the authorities designated under Regulation (EU) 2022/2065 should act as enforcement authorities for the purposes of enforcing this provision.

Recital 41: (41) The fact that an AI system is classified as a high risk AI system under this Regulation should not be interpreted as indicating that the use of the system is necessarily lawful or unlawful under other acts of Union law or under national law compatible with Union law, such as on the protection of personal data, Any such use should continue to occur solely in accordance with the applicable requirements resulting from the Charter and from the applicable acts of secondary Union law and national law.

Change 36

Changed:Recital 41 a (new): (41 a)(41a) A number of legally binding rules at European, national and international level already apply or are relevant to AI systems today, including but not limited to EU primary law (the Treaties of the European Union and its Charter of Fundamental Rights), EU secondary law (such as the General Data Protection Regulation, the Product Liability Directive, the Regulation on the Free Flow of Non-Personal Data, anti-discrimination Directives, consumer law and Safety and Health at Work Directives), the UN Human Rights treaties and the Council of Europe conventions (such as the European Convention on Human Rights), and national law. Besides horizontally applicable rules, various domain-specific rules exist that apply to particular AI applications (such as for instance the Medical Device Regulation in the healthcare sector).

4 unchanged paragraphs

Recital 42: (42) To mitigate the risks from high-risk AI systems placed or otherwise put into service on the Union market for deployers and affected persons, certain mandatory requirements should apply, taking into account the intended purpose, the reasonably foreseeable misuse of the system and according to the risk management system to be established by the provider. These requirements should be objective-driven, fit for purpose, reasonable and effective, without adding undue regulatory burdens or costs on operators.

Recital 43: (43) Requirements should apply to high-risk AI systems as regards the quality and relevance of data sets used, technical documentation and record-keeping, transparency and the provision of information to deployers, human oversight, and robustness, accuracy and cybersecurity. Those requirements are necessary to effectively mitigate the risks for health, safety and fundamental rights, as well as the environment, democracy and rule of law, as applicable in the light of the intended purpose or reasonably foreseeable misuse of the system, and no other less trade restrictive measures are reasonably available, thus avoiding unjustified restrictions to trade.

Recital 44: (44) Access to data of high quality plays a vital role in providing structure and in ensuring the performance of many AI systems, especially when techniques involving the training of models are used, with a view to ensure that the high-risk AI system performs as intended and safely and it does not become a source of discrimination prohibited by Union law. High quality training, validation and testing data sets require the implementation of appropriate data governance and management practices. Training, and where applicable, validation and testing data sets, including the labels, should be sufficiently relevant, representative, appropriately vetted for errors and as complete as possible in view of the intended purpose of the system. They should also have the appropriate statistical properties, including as regards the persons or groups of persons in relation to whom the high-risk AI system is intended to be used, with specific attention to the mitigation of possible biases in the datasets, that might lead to risks to fundamental rights or discriminatory outcomes for the persons affected by the high-risk AI system. Biases can for example be inherent in underlying datasets, especially when historical data is being used, introduced by the developers of the algorithms, or generated when the systems are implemented in real world settings. Results provided by AI systems are influenced by such inherent biases that are inclined to gradually increase and thereby perpetuate and amplify …

Recital 45: (45) For the development and assessment of high-risk AI systems, certain actors, such as providers, notified bodies and other relevant entities, such as digital innovation hubs, testing experimentation facilities and researchers, should be able to access and use high quality datasets within their respective fields of activities which are related to this Regulation. European common data spaces established by the Commission and the facilitation of data sharing between businesses and with government in the public interest will be instrumental to provide trustful, accountable and non-discriminatory access to high quality data for the training, validation and testing of AI systems. For example, in health, the European health data space will facilitate non-discriminatory access to health data and the training of artificial intelligence algorithms on those datasets, in a privacy-preserving, secure, timely, transparent and trustworthy manner, and with an appropriate institutional governance. Relevant competent authorities, including sectoral ones, providing or supporting the access to data may also support the provision of high-quality data for the training, validation and testing of AI systems.

Change 37

Changed:Recital 45 a (new): (45 a)(45a) The right to privacy and to protection of personal data must be guaranteed throughout the entire lifecycle of the AI system. In this regard, the principles of data minimisation and data protection by design and by default, as set out in Union data protection law, are essential when the processing of data involves significant risks to the fundamental rights of individuals. Providers and users of AI systems should implement state-of-the-art technical and organisational measures in order to protect those rights. Such measures should include not only anonymisation and encryption, but also the use of increasingly available technology that permits algorithms to be brought to the data and allows valuable insights to be derived without the transmission between parties or unnecessary copying of the raw or structured data themselvesthemselves.

Recital 46: (46) Having comprehensible information on how high-risk AI systems have been developed and how they perform throughout their lifetime is essential to verify compliance with the requirements under this Regulation. This requires keeping records and the availability of a technical documentation, containing information which is necessary to assess the compliance of the AI system with the relevant requirements. Such information should include the general characteristics, capabilities and limitations of the system, algorithms, data, training, testing and validation processes used as well as documentation on the relevant risk management system. The technical documentation should be kept up to date appropriately throughout the lifecycle of the AI system. AI systems can have a large important environmental impact and high energy consumption during their lifecyle. In order to better apprehend the impact of AI systems on the environment, the technical documentation drafted by providers should include information on the energy consumption of the AI system, including the consumption during development and expected consumption during use. Such information should take into account the relevant Union and national legislation. This reported information should be comprehensible, comparable and verifiable and to that end, the Commission should develop guidelines on a harmonised metholodogy for calculation and reporting of this information. To ensure that a single documentation is possible, term…

Change 38

Changed:Recital 46 a (new): (46 a)(46a) AI systems should take into account state-of-the art methods and relevant applicable standards to reduce the energy use, resource use and waste, as well as to increase their energy efficiency and the overall efficiency of the system. The environmental aspects of AI systems that are significant for the purposes of this Regulation are the energy consumption of the AI system in the development, training and deployment phase as well as the recording and reporting and storing of this data. The design of AI systems should enable the measurement and logging of the consumption of energy and resources at each stage of development, training and deployment. The monitoring and reporting of the emissions of AI systems must be robust, transparent, consistent and accurate. In order to ensure the uniform application of this Regulation and stable legal ecosystem for providers and deployers in the Single Market, the Commission should develop a common specification for the methodology to fulfil the reporting and documentation requirement on the consumption of energy and resources during development, training and deployment. Such common specifications on measurement methodology can develop a baseline upon which the Commission can better decide if future regulatory interventions are needed, upon conducting an impact assessment that takes into account existing law.

Change 39

Changed:Recital 46 b (new): (46 b)(46b) In order to achieve the objectives of this Regulation, and contribute to the Union’s environmental objectives while ensuring the smooth functioning of the internal market, it may be necessary to establish recommendations and guidelines and, eventually, targets for sustainability. For that purpose the Commission is entitled to develop a methodology to contribute towards having Key Performance Indicators (KPIs) and a reference for the Sustainable Development Goals (SDGs). The goal should be in the first instance to enable fair comparison between AI implementation choices providing incentives to promote using more efficient AI technologies addressing energy and resource concerns. To meet this objective this Regulation should provide the means to establish a baseline collection of data reported on the emissions from development and training and for deployment;deployment.

4 unchanged paragraphs

Recital 47 a (new): (47a) Such requirements on transparency and on the explicability of AI decision-making should also help to counter the deterrent effects of digital asymmetry and so-called ‘dark patterns’ targeting individuals and their informed consent.

Recital 49: (49) High-risk AI systems should perform consistently throughout their lifecycle and meet an appropriate level of accuracy, robustness and cybersecurity in accordance with the generally acknowledged state of the art. Performance metrics and their expected level should be defined with the primary objective to mitigate risks and negative impact of the AI system. The expected level of performance metrics should be communicated in a clear, transparent, easily understandable and intelligible way to the deployers. The declaration of performance metrics cannot be considered proof of future levels, but relevant methods need to be applied to ensure consistent levels during use While standardisation organisations exist to establish standards, coordination on benchmarking is needed to establish how these standardised requirements and characteristics of AI systems should be measured. The European Artificial Intelligence Office should bring together national and international metrology and benchmarking authorities and provide non-binding guidance to address the technical aspects of how to measure the appropriate levels of performance and robustness.

Recital 50: (50) The technical robustness is a key requirement for high-risk AI systems. They should be resilient against risks connected to the limitations of the system (e.g. errors, faults, inconsistencies, unexpected situations) as well as against malicious actions that may compromise the security of the AI system and result in harmful or otherwise undesirable behaviour. Failure to protect against these risks could lead to safety impacts or negatively affect the fundamental rights, for example due to erroneous decisions or wrong or biased outputs generated by the AI system. Users of the AI system should take steps to ensure that the possible trade-off between robustness and accuracy does not lead to discriminatory or negative outcomes for minority subgroups.

Recital 51: (51) Cybersecurity plays a crucial role in ensuring that AI systems are resilient against attempts to alter their use, behaviour, performance or compromise their security properties by malicious third parties exploiting the system’s vulnerabilities. Cyberattacks against AI systems can leverage AI specific assets, such as training data sets (e.g. data poisoning) or trained models (e.g. adversarial attacks or confidentiality attacks), or exploit vulnerabilities in the AI system’s digital assets or the underlying ICT infrastructure. To ensure a level of cybersecurity appropriate to the risks, suitable measures should therefore be taken by the providers of high-risk AI systems, as well as the notified bodies, competent national authorities and market surveillance authorities, also taking into account as appropriate the underlying ICT infrastructure. High-risk AI should be accompanied by security solutions and patches for the lifetime of the product, or in case of the absence of dependence on a specific product, for a time that needs to be stated by the manufacturer.

Change 40

Changed:Recital 53 a (new): (53 a)(53a) As signatories to the United Nations Convention on the Rights of Persons with Disabilities (UNCRPD), the Union and the Member States are legally obliged to protect persons with disabilities from discrilmination and promote their equality, to ensure that persons with disabilities have access, on an equal basis wirh others, to information and communications technologies and systems, and to ensure respect for privacy for persons with disabilities. Given the growing importance and use of AI systems, the application of universal design principles to all new technologies and services should ensure full, equal, and unrestricted access for everyone potentially affected by or using AI technologies, including persons with disabilities, in a way that takes full account of their inherent dignity and diversity. It is therefore essential that Providers ensure full compliance with accessibility requirements, including Directive (EU) 2016/2102 and Directive (EU) 2019/882. Providers should ensure compliance with these requirements by design. Therefore, the necessary measures should be integrated as much as possible into the design of the high-risk AI system.

Recital 54: (54) The provider should establish a sound quality management system, ensure the accomplishment of the required conformity assessment procedure, draw up the relevant documentation and establish a robust post-market monitoring system. For providers that have already in place quality management systems based on standards such as ISO 9001 or other relevant standards, no duplicative quality management system in full should be expected but rather an adaptation of their existing systems to certain aspects linked to compliance with specific requirements of this Regulation. This should also be reflected in future standardization activities or guidance adopted by the Commission in this respect. Public authorities which put into service high-risk AI systems for their own use may adopt and implement the rules for the quality management system as part of the quality management system adopted at a national or regional level, as appropriate, taking into account the specificities of the sector and the competences and organisation of the public authority in question.

Recital 56: (56) To enable enforcement of this Regulation and create a level-playing field for operators, and taking into account the different forms of making available of digital products, it is important to ensure that, under all circumstances, a person established in the Union can provide authorities with all the necessary information on the compliance of an AI system. Therefore, prior to making their AI systems available in the Union, providers established outside the Union shall, by written mandate, appoint an authorised representative established in the Union.

Recital 58: (58) Given the nature of AI systems and the risks to safety and fundamental rights possibly associated with their use, including as regards the need to ensure proper monitoring of the performance of an AI system in a real-life setting, it is appropriate to set specific responsibilities for deployers. Deployers should in particular use high-risk AI systems in accordance with the instructions of use and certain other obligations should be provided for with regard to monitoring of the functioning of the AI systems and with regard to record-keeping, as appropriate.

Change 41

Changed:Recital 58 a (new): (58 a)(58a) Whilst risks related to AI systems can result from the way such systems are designed, risks can as well stem from how such AI systems are used. Deployers of high-risk AI system therefore play a critical role in ensuring that fundamental rights are protected, complementing the obligations of the provider when developing the AI system. Deployers are best placed to understand how the high-risk AI system will be used concretely and can therefore identify potential significant risks that were not foreseen in the development phase, due to a more precise knowledge of the context of use, the people or groups of people likely to be affected, including marginalised and vulnerable groups. Deployers should identify appropriate governance structures in that specific context of use, such as arrangements for human oversight, complaint-handling procedures and redress procedures, because choices in the governance structures can be instrumental in mitigating risks to fundamental rights in concrete use-cases. In order to efficiently ensure that fundamental rights are protected, the deployer of high-risk AI systems should therefore carry out a fundamental rights impact assessment prior to putting it into use. The impact assessment should be accompanied by a detailed plan describing the measures or tools that will help mitigating the risks to fundamental rights identified at the latest from the time of putting it into use. If such plan cannot be identified, the deployer should refrain from…from …

Recital 59: (59) It is appropriate to envisage that the deployer of the AI system should be the natural or legal person, public authority, agency or other body under whose authority the AI system is operated except where the use is made in the course of a personal non-professional activity.

Recital 60: (60) Within the AI value chain multiple entities often supply tools and services but also components or processes that are then incorporated by the provider into the AI system, including in relation to data collection and pre-processing, model training, model retraining, model testing and evaluation, integration into software, or other aspects of model development. The involved entities may make their offering commercially available directly or indirectly, through interfaces, such as Application Programming Interfaces (API), and distributed under free and open source licenses but also more and more by AI workforce platforms, trained parameters resale, DIY kits to build models or the offering of paying access to a model serving architecture to develop and train models. In the light of this complexity of the AI value chain, all relevant third parties, in particular those that are involved in the development, sale and the commercial supply of software tools, components, pre-trained models or data incorporated into the AI system, or providers of network services, should without compromising their own intellectual property rights or trade secrets, make available the required information, training or expertise and cooperate, as appropriate, with providers to enable their control over all compliance relevant aspects of the AI system that falls under this Regulation. To allow a cost-effective AI value chain governance, the level of control shall be explicitly disclosed by each third …

Change 42

Changed:Recital 60 a (new): (60 a)(60a) Where one party is in a stronger bargaining position, there is a risk that that party could leverage such position to the detriment of the other contracting party when negotiating the supply of tools, services, components or processes that are used or integrated in a high risk AI system or the remedies for the breach or the termination of related obligations. Such contractual imbalances particularly harm micro, small and medium-sized enterprises as well as start-ups, unless they are owned or sub-contracted by an enterprise which is able to compensate the sub-contractor appropriately, as they are without a meaningful ability to negotiate the conditions of the contractual agreement, and may have no other choice than to accept ‘take-it-or-leave-it’ contractual terms. Therefore, unfair contract terms regulating the supply of tools, services, components or processes that are used or integrated in a high risk AI system or the remedies for the breach or the termination of related obligations should not be binding to such micro, small or medium-sized enterprises and start-ups when they have been unilaterally imposed on them.

Change 43

Changed:Recital 60 b (new): (60 b)(60b) Rules on contractual terms should take into account the principle of contractual freedom as an essential concept in business-to-business relationships. Therefore, not all contractual terms should be subject to an unfairness test, but only to those terms that are unilaterally imposed on micro, small and medium-sized enterprises and start-ups. This concerns ‘take-it-or-leave-it’ situations where one party supplies a certain contractual term and the micro, small or medium-sized enterprise and start-up cannot influence the content of that term despite an attempt to negotiate it. A contractual term that is simply provided by one party and accepted by the micro, small, medium-sized enterprise or a start-up or a term that is negotiated and subsequently agreed in an amended way between contracting parties should not be considered as unilaterally imposed.

Change 44

Changed:Recital 60 c (new): (60 c)(60c) Furthermore, the rules on unfair contractual terms should only apply to those elements of a contract that are related to supply of tools, services, components or processes that are used or integrated in a high risk AI system or the remedies for the breach or the termination of related obligations. Other parts of the same contract, unrelated to these elements, should not be subject to the unfairness test laid down in this Regulation.

Change 45

Changed:Recital 60 d (new): (60 d)(60d) Criteria to identify unfair contractual terms should be applied only to excessive contractual terms, where a stronger bargaining position is abused. The vast majority of contractual terms that are commercially more favourable to one party than to the other, including those that are normal in business-to-business contracts, are a normal expression of the principle of contractual freedom and continue to apply. If a contractual term is not included in the list of terms that are always considered unfair, the general unfairness provision applies. In this regard, the terms listed as unfair terms should serve as a yardstick to interpret the general unfairness provision.

Change 46

Changed:Recital 60 e (new): (60 e)(60e) Foundation models are a recent development, in which AI models are developed from algorithms designed to optimize for generality and versatility of output. Those models are often trained on a broad range of data sources and large amounts of data to accomplish a wide range of downstream tasks, including some for which they were not specifically developed and trained. The foundation model can be unimodal or multimodal, trained through various methods such as supervised learning or reinforced learning. AI systems with specific intended purpose or general purpose AI systems can be an implementation of a foundation model, which means that each foundation model can be reused in countless downstream AI or general purpose AI systems. These models hold growing importance to many downstream applications and systems.

Change 47

Changed:Recital 60 f (new): (60 f)(60f) In the case of foundation models provided as a service such as through API access, the cooperation with downstream providers should extend throughout the time during which that service is provided and supported, in order to enable appropriate risk mitigation, unless the provider of the foundation model transfers the training model as well as extensive and appropriate information on the datasets and the development process of the system or restricts the service, such as the API access, in such a way that the downstream provider is able to fully comply with this Regulation without further support from the original provider of the foundation model.