Data associated with the article: "Exploring the Viability of ChatGPT for Personal Data Anonymization in Government: A Comprehensive Analysis of Possibilities, Risks, and Ethical Implications"

Artificial Intelligence (AI) applications are expected to promote government service delivery and quality, more efficient handling of cases, and bias reduction in decision-making. One potential benefit of the AI tool ChatGPT is that it may support governments in the anonymization of data. However, it is not clear whether ChatGPT is appropriate to support data anonymization for public organizations. Hence, this study examines the possibilities, risks, and ethical implications for government organizations to employ ChatGPT in the anonymization of personal data. We use a case study approach, combining informal conversations, formal interviews, a literature review, document analysis and experiments to conduct a three-step study. First, we describe the technology behind ChatGPT and its operation. Second, experiments with three types of data (fake data, original literature and modified literature) show that ChatGPT exhibits strong performance in anonymizing these three types of texts. Third, an overview of significant risks and ethical issues related to ChatGPT and its use for anonymization within a specific government organization was generated, including themes such as privacy, responsibility, transparency, bias, human intervention, and sustainability. One significant risk in the current form of ChatGPT is a privacy risk, as inputs are stored and forwarded to OpenAI and potentially other parties. This is unacceptable if texts containing personal data are anonymized with ChatGPT. We discuss several potential solutions to address these risks and ethical issues. This study contributes to the scarce scientific literature on the potential value of employing ChatGPT for personal data anonymization in government. In addition, this study has practical value for civil servants who face the challenges of data anonymization in practice including resource-intensive and costly processes.<br>

人工智能（Artificial Intelligence, AI）应用有望提升政务服务的供给水平与服务质量，实现案件处理效率优化，并减少决策环节的偏见。人工智能工具ChatGPT的一项潜在优势在于，可助力政府部门开展数据匿名化工作。然而目前尚无定论，ChatGPT是否适用于公共机构的数据匿名化场景。有鉴于此，本研究针对政府机构采用ChatGPT开展个人数据匿名化工作的可行性、潜在风险及伦理影响展开探讨。本研究采用案例研究方法，结合非正式座谈、正式访谈、文献综述、文档分析与实验，开展三阶段研究工作：其一，阐释ChatGPT的技术原理与运行逻辑；其二，针对三类数据（虚构数据、原始文献与修改后文献）开展的实验结果表明，ChatGPT在对上述三类文本进行匿名化处理时表现优异；其三，梳理了ChatGPT及其在特定政府机构内用于数据匿名化时所涉及的重大风险与伦理议题，涵盖隐私、责任、透明度、偏见、人为干预与可持续性等维度。当前版本的ChatGPT存在一项显著的隐私风险：用户输入内容会被存储并转发至OpenAI及其他潜在第三方。若使用ChatGPT对包含个人数据的文本进行匿名化处理，此类风险将难以被接受。针对上述风险与伦理议题，本研究探讨了若干可行的应对方案。当前学界关于在政务场景中采用ChatGPT实现个人数据匿名化的潜在价值的研究较为匮乏，本研究为此类稀缺的科学文献作出了有益补充。此外，本研究对于在实际工作中面临数据匿名化挑战（包括资源密集型且成本高昂的处理流程）的公务员群体，具备重要的实践参考价值。