法规注册AI应用

would be most useful to industry.We are considering four options:oSingle sector,single regulator:support innovators to bring AI products to the market incollaboration with a single regulator,focusing on only one chosen industrysector.143oMultiple sectors,single regulator:support AI innovators in collaboration with a single regulatorthat is capable of working across multiple industrysectors.144oSingle sector,multiple regulator:establish a sandbox that only operates in one industry sectorbut is capable of supporting AI innovators whose path to market requires interaction with one ormore regulators operating in thatsector.145HM Government Response to Sir Patrick Vallance’s Pro-Innovation Regulation of Technologies Review,HM Treasury,2023.Regulatory Sandbox,ICO,2022; Regulatory Sandbox,FCA,2022.Innovation Hub:Market Insights,FCA.2023; A Sandbox Approach to Regulating High-Risk Artificial IntelligenceApplications,Tuby et al,2021.Inside the regulatory sandbox:effects on fintech funding,Cornelli et al,2020.For an existing example of this type of model see Regulatory Sandbox,FCA,2022.For an existing example of this type of model see Regulatory Sandbox,ICO,2023.For a report on a pilot of this type of model see:Using machine learning in diagnostic services,CQC,2020.A pro-innovation approach to AI regulationoMultiple sectors,multiple regulators:a sandbox capable of operating with one or moreregulators in one or more industry sectors to help AI innovators reach their target market.TheDRCF is piloting a version of thismodel.146

support regulators.We will work with regulators to develop guidance that helps them implementthe principles in a way that aligns with our expectations for how the framework should operate.Existing legal frameworks already mandate and guide regulators' actions.For example,nearlyall regulators are bound by the Regulators'Code116and all regulators – as public bodies – arerequired to comply with the Human RightsAct.117Our proposed guidance to regulators will seekto ensure that when applying the principles,regulators are supported and encouraged to:oAdopt a proportionate approach that promotes growth and innovation by focusing on the risksthat AI poses in a particular context.oConsider proportionate measures to address prioritised risks,taking into account cross-cuttingrisk assessments undertaken by,or on behalf of,government.oDesign,implement and enforce appropriate regulatory requirements and,where possible,integrate delivery of the principles into existing monitoring,investigation and enforcementprocesses.oDevelop joint guidance,where appropriate,to support industry compliance with the principlesand relevant regulatory requirements.Connected & Automated Mobility 2025,Department for Transport,2022.Regulators’ Code,Office for Product Safety and Standards,2014.Human Rights Act,HM Government,1998A pro-innovation approach to AI regulationoConsider how tools for trustworthy AI like assurance techniques and technical standards cansupport regulatory compliance.oEngage proactively and collaboratively with government’s monitoring and evaluation of theframework.A pro-innovation approach to AI regulationCase Study 3.7:What this means for businessesA fictional company,“AI Fairness Insurance Limited”,has delayed the deployment of anew AI application as – under the current patchwork of relevant regulatory requirements –it has been challenging to identify appropriate compliance actions for AI-driven insuranceproducts.Following implementation of the UK’s new pro-innovation framework to regulate AI,wecould expect to see joint guidance produced collaboratively by the Information

[heading4](drawing on the expertise of all relevant regulators).[content]risks.121RationaleTo be effective,this function must be performed centrally,as the whole regulatorylandscape needs to be considered to provide useful guidance to businesses andconsumers on navigating it.This will ensure that businesses and consumers are able tocontribute to the monitoring and evaluation of the framework and its ongoing iteration.This function will help deliver a regulatory regime that is:• Clear:by helping businesses working across sectors to navigate the regulatorylandscape.• Trustworthy:by increasing awareness of the framework and its requirements amongconsumers and businesses.• Collaborative:by educating and raising awareness to empower businesses andconsumers to participate in the ongoing evaluation and iteration of the framework.• Pro-innovation:by enhancing trust,which is shown to increase AI adoption.Horizon scanningActivities• Monitor emerging trends and opportunities in AI development to ensure that theframework can respond to them effectively.Pro-innovation Regulation of Technologies Review:Digital Technologies,HM Treasury,2023.The Centre for Data Ethics and Innovation(CDEI)Public attitudes report states that the public continue to have limitedawareness of AI,with knowledge mainly of low-risk use cases that are already in use but showing low familiarity with morecomplex AI applications.Public expectations for AI governance(transparency,fairness and accountability),Centre for DataEthics and Innovation,2023.A pro-innovation approach to AI regulation• Proactively convene industry,frontier researchers,academia and other keystakeholders to establish how the AI regulatory framework could support the UK’s AIecosystem to maximise the benefits of emerging opportunities whilst continuing to takea proportionate approach to AI risk.• Support the risk assessment function to identify and prioritise new and emerging AIrisks,working collaboratively with industry,academia,global partners,and regulators.RationaleThis function will support horizon-scanning activities in individual regulators but a central