《歐盟人工智能法案綜合 4:標準和合格評定方案.pdf》由會員分享,可在線閱讀,更多相關《歐盟人工智能法案綜合 4:標準和合格評定方案.pdf(25頁珍藏版)》請在三個皮匠報告上搜索。
1、#AIGG24#AIGG24AI Governance AI Governance Global 2024Global 2024Training 6-7 JuneConference 4-5 June#AIGG24#AIGG24#AIGG24#AIGG24Standards and Conformity Assessment Schemes#AIGG24#AIGG24WELCOME AND INTRODUCTIONSAnsgar Koene,Ph.D.EY Global AI Ethics and Regulatory Leader#AIGG24#AIGG24I.Session Outline
2、II.Welcome and Introductions III.Conformity assessmenti.Self&third-party assessmentii.Registration of high-risk systemsiii.Harmonised standardsiv.Common specificationsv.Certificationvi.Derogation from conformity assessmentvii.Codes of Practice for GPAI&systemic risk GPAIviii.Voluntary Codes of Condu
3、ctIV.Questions and AnswersV.Closing RemarksPROGRAM OUTLINE#AIGG24#AIGG24DefinitionsConformity AssessmentAn assessment to ensure that a product,service,process,system or person meets relevant requirements,before it is placed on the market.#AIGG24#AIGG24Conformity assessmentAssessment based on interna
4、l controlsA self-assessment by the provider without the involvement of an approved third party assessment body(notified body)Third party assessmentAssessment by an independent notified body that has been approved to perform third party assessments by a national notifying authority.#AIGG24#AIGG24Self
5、 or third-party assessment?AI application placed-on/used-in EU marketHigh-risk AI application?No obligationAssessment based on internal controls(self-assessment)Third-party assessment(see Annex VII)Component of a system that already undergoes thirds-party assessment?Biometric identification,categori
6、zation or emotion recognition?yesYes(Annex I)yesnoNo(Annex III)noUsing harmonized standards or common specifications?noyes#AIGG24#AIGG24Conformity assessment procedure based on internal controlsThe provider:verifies that the established quality management system is compliant with the requirements of
7、 Article 17verifies the information contained in the technical documentation to assess the compliance of the AI system with the relevant essential requirements for High-Risk AI systemsverifies that the design and development process of the AI system and its post-market monitoring as referred to in A
8、rticle 61 is consistent with the technical documentation#AIGG24#AIGG24Third-part assessment by notified bodyThe notified body:assesses the documentation of the Quality Management Systemand the procedure to ensure the QMS remains adequate and effectivedirectly assesses the QMSexamines the technical d
9、ocumentationmay perform limited examination of training,validation and testing data or request further evidenceThe provider must notify about any significant change to the QRM or AI system that might require a re-assessment.#AIGG24#AIGG24Key information to be submitted by provider for registration o
10、f high-risk AI systemdescription of the intended purpose and of the components and functions supported through this AI systembasic and concise description of the information used by the system(data,inputs)and its operating logicstatus of the AI system(in service;no longer in service,recalled)EU memb
11、er states in which the AI system has been made availablea copy of the EU declaration of conformityelectronic instructions for use(shall not be provided for systems in the areas of law enforcement and migration,asylum and border control management)#AIGG24#AIGG24Key information to be submitted by publ
12、ic authority deploying AI systema summary of the findings of the fundamental rights impact assessmenta summary of the data protection impact assessment carried out#AIGG24#AIGG24Key information to be submitted by providers who conclude their Annex III AI systems are not high-riskAI system trade name
13、and any additional unambiguous reference allowing identification and traceability of the AI systemdescription of the intended purpose of the AI systemshort summary of the grounds for considering the AI system as not high-riskstatus of the AI system(on the market,no longer in service,recalled)EU memb
14、er states in which the AI system has been placed on the market#AIGG24#AIGG24HarmonisedStandardsEuropean standards that have been approved by the European Commission as means to demonstrate that products,services,or processes comply with relevant EU legislation.#AIGG24#AIGG24Requirement for EC to iss
15、ue standardisationrequestthe Commission shall issue standardisation requests covering all requirements of this Regulationthe standardisation request shall also ask for deliverables on reporting and documentation processes to improve AI systems resource performance,such as reduction of energy and oth
16、er resources consumption of the high-risk AI system during its lifecycle,and on energy efficient development of general-purpose AI models#AIGG24#AIGG24Standardization Request to be issued by EC(Nov 2023 draft)risk management systemgovernance and quality of datasetsrecord keeping through logging capa
17、bilitiestransparency and information provisions for usershuman oversightaccuracy specificationsrobustness specificationscybersecurity specificationsquality management system for providers of AI systems,including post-market monitoring processesconformity assessment#AIGG24#AIGG24Standards development
18、 organisations#AIGG24#AIGG24Some ISO/IEC standards candidates identified by CEN-CENELC JTC21ISO/IEC CD 42001.AI management system.ISO/IEC CD TS 6254 Objectives and approaches for explainability of ML models and AI systems.ISO/IEC AWI TS 8200 Controllability of automated artificial intelligence syste
19、msISO/IEC AWI 12792 Transparency taxonomy of AI systemsISO/IEC TR 24027 Bias in AI systems and AI aided decision makingISO/IEC 27xxx series on cybersecurityISO/IEC TR 24368:2022:Overview of ethical and societal concerns.ISO/IEC 38507:2022:Governance implications of the use of artificial intelligence
20、 by organizationsISO/IEC CD 42005.AI system impact assessment.ISO/IEC CD 42006 Requirements for bodies providing audit and certification of artificial intelligencemanagement systems#AIGG24#AIGG24Common SpecificationDetailed practical rules issued by the European Commission that outline how specific
21、types of devices or systems should comply with certain requirements.#AIGG24#AIGG24Reasons for the EC to issue a Common Specificationthe standardization request has not been accepted by any European standardization organisationthe harmonised standards addressing that request are not delivered within
22、the deadline setthe relevant harmonised standards insufficiently address fundamental rights concernsthe harmonised standards do not comply with the request#AIGG24#AIGG24Derogation from conformity assessment procedureAny market surveillance authority may authorise specific high-risk AI systems within
23、 a Member State,for exceptional reasons of public security or the protection of life and health of persons,environmental protection and the protection of key industrial and infrastructural assets.That authorisation shall be for a limited period while necessary conformity assessment procedures are be
24、ing carried out.#AIGG24#AIGG24Transparency obligations for GPAI modelsThe AI Office shall encourage and facilitate the drawing up of codes of practice to facilitate effective implementation of detection and labelling of artificially generated or manipulated content.The Commission is empowered to app
25、rove these codes of practice.If it deems the code is not adequate,the Commission is empowered to specify the common rules for the implementation of those obligations.#AIGG24#AIGG24Codes of Practice for systemic risk GPAIThe AI Office and the AI Board shall aim to ensure that the codes of practice co
26、ver:content used for trainingtype and nature of systemic riskassessment and management of systemic risk,including documentationneeds and interest of all affected parties#AIGG24#AIGG24Voluntary Codes of Conduct for non high-risk AI systemselements foreseen in European ethic guidelines for trustworthy
27、 AIassessing and minimizing the impact of AI systems on environmental sustainability promoting AI literacyfacilitating an inclusive and diverse design of AI systemsassessing and preventing negative impact of AI systems on vulnerable persons or groups of persons,including as regards accessibility for
28、 persons with a disability,as well as on gender equality#AIGG24#AIGG24AI Act enforcement timeline linked to time needed to develop codes and standardsAI Act entry into forceimplementation period beginsExpected June 2024Most other AI Act obligations take effect(e.g.,requirements for high-risk systems
29、 described in Annex III)after 24 months202420252026AI Act prohibitions take effectafter 6 months AI Act GPAI obligations take effectafter 12 months2027AI Act obligations take effect for high-risk systems that are part of union harmonized safety legislation(Annex I)after 36 months2028High-risk AI sys
30、tems used by public authorities on the market before AI Act entry into force to be compliantDecember 2030#AIGG24#AIGG24Did you enjoy this session?Is there any way we could make it better?Let us know by filling out a speaker evaluation.1.Open the Cvent Events app.2.Enter IAPP AIGG24(case and space sensitive)in search bar.3.Tap Schedule on the bottom navigation bar.4.Find this session.Click Rate this Session within the description.5.Once youve answered all three questions,tap Done.How did things go?(We really want to know)