EU AI Act |

Part 1: Regulated Product or Safety Component High-Risk AI

An AI system is high-risk where:

• It is a product covered by the legislation listed in Annex I of the Act (summarized in the boxes below) that is required by that legislation to undergo a third-party conformity assessment; or
• It is a safety component of such a product.

^{Radio Equipment}

Directive 2014/53/EU of the European Parliament and of the Council of 16 April 2014 on the harmonisation of the laws of the Member States relating to the making available on the market of radio equipment and repealing Directive 1999/5/EC (OJ L 153, 22.5.2014, p. 62).

^{Cableway Installations}

Regulation (EU) 2016/424 of the European Parliament and of the Council of 9 March 2016 on cableway installations and repealing Directive 2000/9/EC (OJ L 81, 31.3.2016, p. 1).

^Lifts

Directive 2014/33/EU of the European Parliament and of the Council of 26 February 2014 on the harmonisation of the laws of the Member States relating to lifts and safety components for lifts (OJ L 96, 29.3.2014, p. 251).

^Machinery

Directive 2006/42/EC of the European Parliament and of the Council of 17 May 2006 on machinery, and amending Directive 95/16/EC (OJ L 157, 9.6.2006, p. 24).

^{Safety of Toy}

Directive 2009/48/EC of the European Parliament and of the Council of 18 June 2009 on the safety of toys (OJ L 170, 30.6.2009, p. 1).

^{Gas Appliances}

Regulation (EU) 2016/426 of the European Parliament and of the Council of 9 March 2016 on appliances burning gaseous fuels and repealing Directive 2009/142/EC (OJ L 81, 31.3.2016, p. 99).

^{Pressure Equipment}

Directive 2014/68/EU of the European Parliament and of the Council of 15 May 2014 on the harmonisation of the laws of the Member States relating to the making available on the market of pressure equipment (OJ L 189, 27.6.2014, p. 164).

^{Explosive Atmospheres}

Directive 2014/34/EU of the European Parliament and of the Council of 26 February 2014 on the harmonisation of the laws of the Member States relating to equipment and protective systems intended for use in potentially explosive atmospheres (OJ L 96, 29.3.2014, p. 309).

^{Medical Devices}

Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices, amending Directive 2001/83/EC, Regulation (EC) No 178/2002 and Regulation (EC) No 1223/2009 and repealing Council Directives 90/385/EEC and 93/42/EEC (OJ L 117, 5.5.2017, p. 1).

^{In Vitro Medical Devices}

Regulation (EU) 2017/746 of the European Parliament and of the Council of 5 April 2017 on in vitro diagnostic medical devices and repealing Directive 98/79/EC and Commission Decision 2010/227/EU (OJ L 117, 5.5.2017, p. 176).

^{Protective Equipment}

Regulation (EU) 2016/425 of the European Parliament and of the Council of 9 March 2016 on personal protective equipment and repealing Council Directive 89/686/EEC (OJ L 81, 31.3.2016, p. 51).

^{Marine Equipment}

Directive 2014/90/EU of the European Parliament and of the Council of 23 July 2014 on marine equipment and repealing Council Directive 96/98/EC (OJ L 257, 28.8.2014, p. 146).

^{Motor Vehicles}

Regulation (EU) 2018/858 of the European Parliament and of the Council of 30 May 2018 on the approval and market surveillance of motor vehicles and their trailers, and of systems, components and separate technical units intended for such vehicles, amending Regulations (EC) No 715/2007 and (EC) No 595/2009 and repealing Directive 2007/46/EC (OJ L 151, 14.6.2018, p. 1).

^{Motor Vehicles}

Regulation (EU) 2019/2144 of the European Parliament and of the Council of 27 November 2019 on type-approval requirements for motor vehicles and their trailers, and systems, components and separate technical units intended for such vehicles, as regards their general safety and the protection of vehicle occupants and vulnerable road users, amending Regulation (EU) 2018/858 of the European Parliament and of the Council and repealing Regulations (EC) No 78/2009, (EC) No 79/2009 and (EC) No 661/2009 of the European Parliament and of the Council and Commission Regulations (EC) No 631/2009, (EU) No 406/2010, (EU) No 672/2010, (EU) No 1003/2010, (EU) No 1005/2010, (EU) No 1008/2010, (EU) No 1009/2010, (EU) No 19/2011, (EU) No 109/2011, (EU) No 458/2011, (EU) No 65/2012, (EU) No 130/2012, (EU) No 347/2012, (EU) No 351/2012, (EU) No 1230/2012 and (EU) 2015/166 (OJ L 325, 16.12.2019, p. 1).

^{Motorcycles / Quads}

Regulation (EU) No 168/2013 of the European Parliament and of the Council of 15 January 2013 on the approval and market surveillance of two- or three-wheel vehicles and quadricycles (OJ L 60, 2.3.2013, p. 52).

^{Rail System}

Directive (EU) 2016/797 of the European Parliament and of the Council of 11 May 2016 on the interoperability of the rail system within the European Union (OJ L 138, 26.5.2016, p. 44).

^{Farm / Forest Vehicles}

Regulation (EU) No 167/2013 of the European Parliament and of the Council of 5 February 2013 on the approval and market surveillance of agricultural and forestry vehicles (OJ L 60, 2.3.2013, p. 1).

^{Crafts / Watercrafts}

Directive 2013/53/EU of the European Parliament and of the Council of 20 November 2013 on recreational craft and personal watercraft and repealing Directive 94/25/EC (OJ L 354, 28.12.2013, p. 90).

^{Civil Aviation Rules}

Regulation (EU) 2018/1139 of the European Parliament and of the Council of 4 July 2018 on common rules in the field of civil aviation and establishing a European Union Aviation Safety Agency, and amending Regulations (EC) No 2111/2005, (EC) No 1008/2008, (EU) No 996/2010, (EU) No 376/2014 and Directives 2014/30/EU and 2014/53/EU of the European Parliament and of the Council, and repealing Regulations (EC) No 552/2004 and (EC) No 216/2008 of the European Parliament and of the Council and Council Regulation (EEC) No 3922/91 (OJ L 212, 22.8.2018, p. 1), in so far as the design, production and placing on the market of aircrafts referred to in Article 2(1), points (a) and (b) thereof, where it concerns unmanned aircraft and their engines, propellers, parts and equipment to control them remotely, are concerned.

^{Civil Aviation Security}

Regulation (EC) No 300/2008 of the European Parliament and of the Council of 11 March 2008 on common rules in the field of civil aviation security and repealing Regulation (EC) No 2320/2002 (OJ L 97, 9.4.2008, p. 72).

Part 2: Designated High-Risk AI System

All AI systems identified in Annex III of the Act are considered high-risk by default absent an exception / derogation. Below are details. Note Annex III is subject to change.

AI systems intended for identifying individuals without their active involvement, typically at a distance, through the comparison of an individual’s biometric data with the biometric data contained in a reference database.

AI systems intended to be used for biometric categorization, according to sensitive or protected attributes or characteristics based on the inference of those attributes or characteristics.

AI systems intended to be used for emotion recognition.

AI systems intended to be used as safety components in the management and operation of critical digital infrastructure, road traffic, or in the supply of water, gas, heating or electricity.

AI systems intended to be used to:

• Determine access or admission or to assign individuals to educational and vocational training institutions
• Evaluate learning outcomes, including when those outcomes are used to steer the learning process of individuals in educational and vocational training institutions
• Assess the appropriate level of education that an individual will receive or will be able to access, in the context of or within educational and vocational training institutions
• Monitor and detect prohibited behavior of students during tests in the context of or within educational and vocational training institutions

AI systems intended to be used to:

• Recruit or select individuals, including to place targeted job advertisements, analyze and filter job applications and to evaluate candidates
• Make decisions affecting terms of work-related relationships, or the promotion or termination of work-related contractual relationships
• Allocate tasks based on individual behavior or personal traits or characteristics of persons in work-related relationships
• Monitor and evaluate the performance and behavior of persons in work-related relationships

AI systems intended to be used for risk assessment and pricing in relation to individuals in the case of life and health insurance.

AI systems intended to:

• Be used by public authorities or on their behalf to evaluate the eligibility of individuals for essential public assistance benefits and services, including healthcare services, as well as to grant, reduce, revoke, or reclaim such benefits and services
• Be used to evaluate the creditworthiness of individuals or establish their credit score, with the exception of AI systems used for the purpose of detecting financial fraud
• Evaluate and classify emergency calls by individuals or to be used to dispatch, or to establish priority in the dispatching of, emergency first response services, including by police, firefighters and medical aid, as well as of emergency healthcare patient triage systems

AI systems intended to be used by or on behalf of law enforcement authorities, or by Union institutions, bodies, offices or agencies in support of law enforcement authorities:

• As polygraphs or similar tools
• To assess the risk of an individual becoming the victim of criminal offences
• To evaluate the reliability of evidence in the course of the investigation or prosecution of criminal offences
• To assess the risk of an individual offending or re-offending not solely on the basis of the profiling of natural persons as referred to in Article 3(4) of the Data Protection Law Enforcement Directive
• To assess personality traits and characteristics or past criminal behavior of individuals or groups
• To profile natural persons as referred to in the Data Protection Law Enforcement Directive in the course of the detection, investigation or prosecution of criminal offences

AI systems intended to be used by or on behalf of competent public authorities or by Union institutions, bodies, offices or agencies:

• As polygraphs or similar tools
• To assess a risk, including a security risk, a risk of irregular migration, or a health risk, posed by an individual who intends to enter or who has entered into the territory of a Member State
• To assist competent public authorities for the examination of applications for asylum, visa or residence permits and for associated complaints with regard to the eligibility of the individuals applying for a status, including related assessments of the reliability of evidence
• In the context of migration, asylum or border control management, for the purpose of detecting, recognizing or identifying individuals, with the exception of the verification of travel documents

AI systems intended to be used by a judicial authority or on their behalf to assist a judicial authority in researching and interpreting facts and the law and in applying the law to a concrete set of facts, or to be used in a similar way in alternative dispute resolution.

AI systems intended to be used for influencing the outcome of an election or referendum or the voting behavior of individuals in the exercise of their vote in elections or referenda. This does not include AI systems to the output of which individuals are not directly exposed, such as tools used to organize, optimize or structure political campaigns from an administrative or logistical point of view.

Providers of high-risk AI systems must:

• Establish, implement, document and maintain risk management and post-market monitoring systems;
• Establish data quality, management, and governance systems;
• Design and develop the AI system to achieve accuracy, robustness, and security;
• Design the AI system to comply with accessibility requirements and allow for effective human oversight of the AI system;
• Draw-up and keep technical documentation and automatic recording of logs;
• Ensure the AI system undergoes a conformity assessment procedure prior to being placed on the market or put into service, and draw up an appropriate EU declaration of conformity;
• Register the system in the high-risk AI system EU database and affix the CE marking before sale / use (along with the provider’s name, registered trade name or trademark, and the address at which they can be contacted);
• Provide information to users of the AI system (including instructions for use) and maintain appropriate transparency to facilitate compliance;
• Appoint an authorized representative in the EU, if the provider is based outside the EU;
• Immediately inform the competent authority of any serious incident of an AI system;
• Upon request, demonstrate conformity of the AI system to the regulatory requirements; and
• Take immediate corrective actions if the AI system does not comply with law.

Other parties (like distributors, importers or deployers) assume the same, extensive obligations as providers in relation to “high-risk” AI systems, if they: (i) use the AI system under their name or trademark, (ii) modify the intended purpose of the AI system, or (iii) make a substantial modification to the AI system.

Importers of high-risk AI systems must verify that:

• The relevant conformity assessment procedure has been carried out by the provider of the high-risk AI system;
• The provider has drawn up all required technical documentation;
• The AI system bears the required CE marking and is accompanied by the EU declaration of conformity and instructions for use; and
• The provider has appointed an EU representative, where required.

Importers must also:

• Indicate their name, registered trade name or trademark, and the address at which they can be contacted on the high-risk AI system and on its packaging or its accompanying documentation, where applicable;
• Ensure the AI system does not come out of compliance while under their responsibility; and
• Keep appropriate records and, upon request, demonstrate conformity of the AI system to the regulatory requirements.

Distributors of high-risk AI systems must verify that:

• The system bears the required CE marking and is accompanied by the EU declaration of conformity and instructions for use; and
• The provider and importer of the AI system, as applicable, have complied with their respective obligations.

Distributors must also:

• Ensure the AI system does not come out of compliance while under their responsibility;
• Withdraw or take corrective actions to fix an AI system that is not in compliance with the Act’s obligations; and
• Keep appropriate records and, upon request, demonstrate conformity of the AI system to the regulatory requirements.

Deployers of high-risk AI systems must:

• Take appropriate technical and organizational measures to ensure the system is used in accordance with the instruction for use accompanying the system and monitor the operation of the system;
• Implement competent human oversight to the extent the deployer exercises control over the system;
• Ensure relevant and appropriate robustness and cybersecurity measures are regularly monitored for effectiveness and are regularly adjusted or updated;
• Ensure that input data is relevant and sufficiently representative to the extent the deployer exercises control over the input data;
• Maintain the logs automatically generated by the AI system to the extent they are under their control;
• Consult workers’ representatives and inform the affected employees that they will be subject to the system prior to putting into service or using a high-risk AI system at the workplace;
• Inform people that they are subject to the use of the high-risk AI system and their right to an explanation, if high-risk AI systems are being used to make decisions or assist in making decisions related to natural persons; and
• Perform an assessment of the system’s impact in the specific context of its use.

An AI model:

• That displays significant generality; and
• Is capable of competently performing a wide range of distinct tasks; and
• That can be integrated into a variety of downstream systems or applications.

AI models with at least a billion parameters trained with a large amount of data using self-supervision at scale should generally be considered as displaying significant generality and competence to perform a wide range of distinctive tasks.

A general-purpose AI model meeting any of the following criteria will be classified as presenting systemic risk:

• It has high impact capabilities evaluated on the basis of appropriate technical tools and methodologies, including indicators or benchmarks (which is presumed where the cumulative amount of compute used for its training measured in floating point operations (FLOPs) is greater than 10^25); or
• The European Commission decides the model has capabilities or impact equivalent to those identified above based on the criteria listed in Annex XIII, including number of parameters of the model; quality or size of the data set (e.g., measured through tokens); the amount of compute used for training the model, measured in FLOPS or using other variables (e.g., estimated cost of training, time required for training or energy consumption); input and output modalities; benchmarks and evaluations of capabilities of the model; high impact on the internal market due to reach (presumed when it has been made available to at least 10k registered business users established in the EU); and number of registered end-users.

Providers of General-Purpose AI Models Must:

• Create and keep updated technical documentation of the model (elements set out in Annex XI), to be provided upon request to the AI Office and national competent authorities (unless the model is released under a free and open-source license meeting certain criteria)
• Create, keep updated, and make available information and documentation of the model (elements set out in Annex XII) to providers of AI systems intending to integrate the model into their AI systems (unless the model is released under a free and open-source license meeting certain criteria)
• Put in place a policy to respect EU copyright laws and in particular identify and respect reservation of rights regarding scraping in the Copyright Directive
• Create and make publicly available a summary of the content used for training the model according to a template to be provided by the AI office

Providers of General-Purpose AI Models presenting systemic risk must also:

• Conduct a classification exercise and notify the AI office that the model qualifies as a general-purpose AI model with systemic risk
• Perform model evaluations in accordance with standardized protocols and tools reflecting the state of the art
• Assess and mitigate possible systemic risks that may stem from the development, the placing on the market, or the use of the model
• Track, document and report information about serious incidents and corrective measures without undue delay to the AI Office, and, as appropriate, to national competent authorities
• Ensure an adequate level of cybersecurity protection for the model and its physical infrastructure

• If an AI system will interact directly with individuals, the system must be designed and developed to ensure the individual is informed that they are interacting with an AI system (unless it is obvious based on the context)
• If an AI system performs emotion recognition or biometric categorization, individuals exposed to the system must be informed of the operation of the system
• If an AI system generates text that is used for the purpose of informing the public on matters of public interest, it must be disclosed that the text has been artificially generated or manipulated
• If an AI system generates synthetic content, the system’s outputs must be marked in a machine-readable and detectable format as artificially generated or manipulated.
• If an AI system generates deep fakes, it must be disclosed that the content has been manipulated or artificially generated