Accelerating Trustworthy AI: developing a first-to-market AI System Risk Management Platform for Insurance Product creation

加速可信人工智能：开发首个上市的人工智能系统风险管理平台，用于保险产品创建

• 批准号: 10093285
• 金额: $ 104.97万
• 依托单位: HOLISTIC AI LIMITED
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2024
• 资助国家: 英国
• 起止时间: 2024 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=10093285
• 关键词: Accelerating; Trustworthy; AI; developing; first

Algorithm adoption is rapidly expanding across business and society. The number of AI companies in the UK has grown by 600% over the last ten years to \>1,300 companies, and AI adoption within businesses is predicted to reach 1.3m by 2040 (UK Gov).But as AI grows, so do concerns around potential consequences of mass adoption. Legislation is accelerating globally: The UK has announced its National AI Strategy including the 'assurance ecosystem roadmap'; while the far-reaching EU-AI-Act and multiple US interventions (e.g. New York City AI-HR 'bias audits') are emerging and already dramatically changing the landscape. This regulation is sorely needed. Soon billions of algorithms will make decisions with minimal human intervention; from autonomous vehicles and finance, to medical treatment, employment, and legal decisions. Without protection against bias, errors, and other potentially damaging and costly issues, companies face major losses and potential fines.Following our Phase 1 Accelerating Trustworthy AI project, Holistic AI (HAI) and the wider consortium is ideally positioned to respond to this major and still growing industry need for assessment of companies' AI position and insurance against potential losses. We sit at the intersection between AI assurance and insurance services, a crucial balance for solving of this issue, and our team is uniquely experienced in both sides of the equation. We currently work with enterprise companies like Unilever, evaluating their multiple AI projects for bias and failure, ready for multinational application in the emerging regulatory landscape.We now seek to create a full-service Risk Management Platform, building on our work during Phase 1, to deliver on insurance industry requirements. The platform will enable insurers to rapidly evaluate and quantify potential risk within and from AI and data applications to provide customers with reliable, accurate, insurance quotes. We will leverage our existing assessment capabilities, alongside vital datasets gathered during development to-date and industry partner data, case-studies and knowledge to build a pricing and assessment structure fundamental to enable the appropriate insurance of algorithms and their deployment.This project will therefore create a unique and novel offering providing the first future-proof tools suitable for insurer-based AI assessment, answering a major industry demand and a commercial necessity for the maturity of trusted AI deployment. It creates a significant global market innovation centralised in the UK, answering both national priorities around AI growth and responsible deployment, and growing global concerns around ethical AI usage.

算法的采用正在企业和社会中迅速扩大。英国的人工智能公司数量在过去 10 年里增长了 600%，达到 1,300 家以上，预计到 2040 年，企业内部人工智能的采用将达到 130 万家（英国政府）。但随着人工智能的发展，人们的担忧也在增加围绕大规模采用的潜在后果。全球立法正在加速：英国宣布了国家人工智能战略，包括“保障生态系统路线图”；而影响深远的《欧盟人工智能法案》和美国的多项干预措施（例如纽约市人工智能人力资源“偏见审计”）正在兴起，并已经极大地改变了这一格局。这项规定是非常需要的。很快，数十亿算法将在最少的人为干预下做出决策；从自动驾驶汽车和金融，到医疗、就业和法律决策。如果不能防止偏见、错误和其他潜在破坏性和代价高昂的问题，公司将面临重大损失和潜在的罚款。在我们的第一阶段加速可信人工智能项目之后，整体人工智能 (HAI) 和更广泛的联盟处于理想的位置来应对这一重大和行业对公司人工智能地位评估和潜在损失保险的需求仍在不断增长。我们处于人工智能保障和保险服务之间的交叉点，这是解决这一问题的关键平衡点，我们的团队在这两方面都拥有独特的经验。我们目前与联合利华等企业合作，评估他们的多个人工智能项目是否存在偏见和失败，为新兴监管环境中的跨国应用做好准备。我们现在寻求在第一阶段工作的基础上创建一个全方位服务的风险管理平台，以满足保险业的要求。该平台将使保险公司能够快速评估和量化人工智能和数据应用程序内部的潜在风险，为客户提供可靠、准确的保险报价。我们将利用我们现有的评估能力，以及迄今为止开发过程中收集的重要数据集和行业合作伙伴数据、案例研究和知识，构建一个定价和评估结构的基础，以实现算法及其部署的适当保障。因此，该项目将创建独特而新颖的产品，提供第一个适合基于保险公司的人工智能评估的面向未来的工具，满足主要行业需求和可信人工智能部署成熟的商业需求。它创造了一个以英国为中心的重要全球市场创新，既解决了围绕人工智能增长和负责任部署的国家优先事项，也解决了全球对人工智能使用道德日益增长的担忧。