21世纪经济报道记者 黄子潇 深圳报道 中银香港副总裁陈文表示，该行已于首期GenA.I.沙盒计划取得成果，本次再次入选。该行将于第二期 通过六个月的概念验证，探索GenA.I.在财富管理、风险管理等领域的应用潜力。 历时半年申请，香港金管局日前公布第二期生成式人工智能（GenA.I.）沙盒参与者名单。在超60个方 案中，20家银行和14家技术供应商的27个用例入选第二期沙盒名单。 此前2024年12月，该局曾发布首批沙盒名单，10家银行和四家技术供应商的15个用例入选。 21世纪经济报道记者注意到，目前已有中银香港、招商永隆银行、兴业银行香港分行等中资银行，以及 汇丰、渣打等外资银行入选沙盒，亦有PAO bank（原平安壹账通银行）、蚂蚁银行等数字银行名列其 中。 值得关注的是，相较于首批AI沙盒的初步探索，第二批AI沙盒的多个用例采用了"以AI对抗AI"策略。 记者在多家机构中找到了具体用例。 多家中资银行入选第二批沙盒 香港金管局披露信息显示，第二期沙盒中名单包括20家银行，分别为蚂蚁银行（香港）、中银香港、交 银香港、东亚银行、建银亚洲、集友银行、招商永隆银行、大新银行、恒生银行、富邦银行（中国）、 ...

Core Insights - PAObank, a subsidiary of Ping An Group, has developed an Anti-Fraud Strategy Platform in collaboration with Financial One Account, which has been selected for the Hong Kong Monetary Authority's GenA.I. sandbox, marking a significant advancement in AI application within the banking sector [2][3] - The GenA.I. sandbox, set to launch in 2024, focuses on risk management, anti-fraud measures, and customer experience innovation in the banking industry, particularly addressing deepfake fraud [2][3] Group 1 - The Anti-Fraud Strategy Platform aims to enhance the identification of deepfake technology through AI facial recognition, allowing real-time verification of user selfies to detect potential fraud [3] - PAObank's CTO highlighted the continuous enhancement of AI applications to improve operational efficiency and customer experience, with total customer deposits exceeding 8 billion HKD by Q3 2025 [3][4] - Financial One Account has launched an intelligent visual anti-fraud strategy platform, which has successfully intercepted over 20,000 black market attacks and is expected to cover over 70% of retail banking users in Hong Kong through partnerships with local banks [4]

Core Insights - PAObank, a digital bank under Ping An Group, has successfully launched an "Anti-Fraud Strategy Platform" in collaboration with OneConnect, marking a significant advancement in AI technology application within Hong Kong's banking sector [1][2] - The platform aims to combat deepfake fraud risks by utilizing AI to enhance fraud detection and risk management capabilities in the banking industry [1][3] Group 1: AI Technology and Fraud Prevention - The "Anti-Fraud Strategy Platform" is part of the GenA.I. sandbox launched by the Hong Kong Monetary Authority, focusing on risk management and fraud prevention in the banking sector [1][2] - The platform employs advanced AI facial recognition technology to verify user identities in real-time, helping to detect potential fraudulent activities [2][3] - The platform has a detection defense rate exceeding 99%, showcasing its effectiveness in identifying complex attacks such as face-swapping and forged videos [3] Group 2: Business Performance and Growth - As of Q3 2025, PAObank's total customer deposits have surpassed 8 billion HKD, indicating strong growth in personal banking services [2] - Ping An Group has developed a comprehensive database with over 3.2 trillion text data, 310,000 hours of annotated voice data, and over 7.5 billion image data, enhancing its AI capabilities across various business scenarios [4] - The implementation of AI has led to significant operational efficiencies, including a 94% instant underwriting rate for life insurance policies and a 64.4 billion CNY reduction in claims fraud for property insurance [4]

Core Insights - In 2024, the focus on large model security will shift towards the security of the models themselves, while by 2025, the risks will escalate to include the security of the intelligent agents and their application environments [1] - The rapid advancement of AI technology is driving a transformation across various industries, with predictions indicating that by 2026, 50% of China's top 500 companies will utilize AI agents for data preparation and analysis [1] - The emergence of AI agents has led to a new phase in cybersecurity, termed "intelligent agent confrontation," where cybersecurity firms are developing dynamic defense systems that utilize AI to counter AI threats [2] Company Developments - Qimingxingchen has launched the "Nine Sky. Taihe" security model and the "Anxing" intelligent agent, enhancing their capabilities to identify and manage data risks [2] - The accuracy of data risk identification has improved from 70% to 90% through the integration of AI technologies [2] - The company is actively working on a cluster of intelligent agents that can collaboratively address complex cybersecurity threats, improving operational efficiency and reducing false alarms [3] Strategic Collaborations - Qimingxingchen is collaborating with China Mobile to develop a next-generation defense system, focusing on AI security governance and integrated security solutions [6] - The partnership aims to create a robust security foundation for the digital transformation of China, addressing the challenges posed by the widespread application of large models [6] - The company is committed to continuous product iteration and technological upgrades to establish a comprehensive security protection system for large model applications [6]