国广清科技术总监杨森：应用隐私计算解决ChatGPT隐私泄露风险
采写：管瑞（实习生、中央戏剧学院本科生）
英译：粟文捷（实习生、天津大学本科生）
OpenAI开发的人工智能应用ChatGPT自2022年11月发布以来，持续受到了全球的广泛瞩目，甚至被认为开启了第四次工业革命。
An artificial intelligence program called ChatGPT, created by OpenAI, has attracted substantial global attention since its introduction in November 2022 and is even viewed as signaling the start of the Fourth Industrial Revolution.
而在2023年3月22日，ChatGPT却被曝出存在用户隐私漏洞，用户能够看到其他用户对话历史记录的标题，引发了公众对ChatGPT隐私泄露风险的担忧。3月31日，意大利个人数据保护局宣布禁止使用ChatGPT，德国等其他欧洲国家也陆续跟进发声，表示会考虑禁止ChatGPT收集数据。
On March 22, 2023, it was discovered that OpenAI's ChatGPT had a privacy flaw that allowed users to see the titles of other users' chat histories. Public worries about ChatGPT's privacy dangers were raised by this finding. The usage of ChatGPT was outlawed on March 31 by the Italian Data Protection Authority, and other European nations, including Germany, soon followed suit by stating their intention to do the same.
近年来，数据安全问题成为社会焦点，数据泄露、滥用等数据安全事件频繁发生，人工智能技术在信息泄露等方面的社会性风险，则因为其使用的庞大数据规模而被进一步放大。如何在人工智能模型训练、智能化应用发展的同时兼顾数据安全，使得人工智能产品满足安全合规要求，成为业界持续关注的热点问题，隐私计算技术因其能够提供隐私安全条件下的联邦学习等机制而受到重点关注。
Due to the frequent incidence of data breaches, abuse, and other data security events, data security has recently emerged into the public eye. The enormous amount of data that artificial intelligence technology use increases the societal concerns they pose, such as information leaking. Data security is a big subject in the market right now as we advance AI model training and intelligent apps to assure compliance. Due to its capacity to offer processes like federated learning in privacy-secured conditions, privacy-preserving technologies like privacy computing have attracted a lot of attention.
隐私计算技术融合了人工智能、密码学、数据科学等众多领域，通过结合安全多方计算、联邦学习、同态加密、差分隐私和机密计算等为代表的现代密码学和信息安全技术，能够在保护数据本身不对外泄露的前提下，实现对数据处于加密状态或非透明状态下的计算和分析，达到对数据“可用、不可见”的目的。
Data science, artificial intelligence, and cryptography are all integrated into privacy computing. Modern information security and cryptography techniques, such as secure multiparty computation, federated learning, homomorphic encryption, differential privacy, and confidential computing, can be combined to create privacy computing, which enables computations on data while ensuring that it is shielded from outside disclosure. It accomplishes the task of making data "usable but invisible" by preserving the data in an encrypted or opaque form without disclosing it to the public.
AI大数据时代，原生互联网公司积累了大量数字技术，并且正在助力千行百业进行数字化转型。数据要素在其中主要有两方面作用，一个是创造，帮助传统企业构建新的模式，进而衍生出新的业态；另一方面，数据要素也在放大其它的生产要素，驱动人力、资本、土地发挥更大的价值。数据要素对人工智能技术健康发展起着至关重要的作用,为推动城市智能化升级和经济可持续发展提供创新实践的源动力。
Native internet firms have amassed a large quantity of digital technology in the age of AI and big data, which is being used to accelerate the digital transformation of several sectors. In this procedure, data pieces primarily play two roles. They do this in two ways. First, they open up new prospects by assisting established businesses in establishing creative business models. Second, data elements boost the productivity of other elements like labor, capital, and land so they can produce more value. Data components are essential to the healthy advancement of artificial intelligence technology, acting as the impetus for creative strategies that support improvements in urban intelligence and long-term economic growth.
传统机构数字化转型具有全面性，不仅仅是技术层面的运用，其重点主要表现在六个方面：适应数字经济的企业文化，敏捷、稳定的技术架构，完善的数据资产化能力，高效的数字化营销和运营能力，健全的数字化风险管控能力以及开放的数字化生态构建能力。人工智能的场景应用在经历技术创新、普及再到变革之后，将迎来一个蓬勃发展的新时代，信息孤岛、基础设施安全、技术可控性等挑战也在逐步显现，致力于实现多方数据“可用不可见”的隐私计算技术，成为关键的技术解决之道。
Beyond only using technology, conventional institutions are also undergoing a digital transition. It primarily focuses on six key areas: fostering an organization culture that can adapt to the digital economy, establishing agile and reliable technological architectures, enhancing data monetization capabilities, optimizing digital marketing and operational efficiency, putting in place strong digital risk management capabilities, and creating an open digital ecosystem. Artificial intelligence's use scenarios has experienced technical advancement, wide adoption, and dramatic transformations, ushering in a prosperous new era. Information silos, infrastructure security, and technical control are problems that have steadily surfaced. The "usability without visibility" of multi-party data is what privacy computing technology, which strives to achieve, achieves, becomes a critical technical answer in resolving these issues.
人工智能作为这个时代最具影响力的技术进步，已经在逐步改变全球经济的方方面面，随着人工智能技术不断取得突破，人类社会将逐渐迈入强人工智能阶段，而隐私计算技术也将作为人工智能模型的重要安全训练手段快速成长并驱动人工智能应用发展。
The most significant technical development of this time period, artificial intelligence, is progressively changing many facets of the world economy. The development of AI technology is leading to a stronger artificial intelligence age for human society. In this context, the growth of AI applications is being fueled by the privacy computing technology, which is quickly developing as a crucial security training technique for AI models.
同时，我国数字经济“十四五”规划强调强调了充分发挥数据要素价值的必要性，隐私计算技术能够为充分发挥海量数据和丰富应用场景优势，有力促进数字技术与经济社会发展各领域融合发展，加快实现数字化发展、建设数字中国的远景目标提供重要技术基础，成为数字化时代激发数据要素价值的利器。
The "14th Five-Year Plan" for China's digital economy highlights the need of completely unlocking the value of data elements at the same time. Utilizing the benefits of big data and a variety of application scenarios, privacy computing technology may help integrate and advance digital technology across a range of economic and social development domains. As a potent catalyst for enhancing the value of data components in the digital age, it acts as a fundamental technical basis for accelerating digital transformation and the creation of a digital China.