Russia’s efforts to interfere in the 2016 and 2020 U.S. presidential elections were pretty low-tech. Relying on generic bot messaging, low-quality content and mass targeting, their operations probably had limited impact.

俄罗斯干预2016年、2020年美国总统大选的手段技术含量相当低。他们靠一般的机器人账号向大众投放数量极多的即时短信和低质量内容，这些行动的影响可能有限。

Those days are over. With the exponential rise of generative A.I. systems, the greatest danger is no longer a flood of invective and falsehoods on social media. Rather, it is the slow, subtle and corrosive manipulation of online communication — propaganda designed not to shock, but to slip silently into our everyday digital discussions. We have entered a new era in international influence operations, where A.I.-generated narratives shift the political landscape without drawing attention.

那些日子已是过去式。随着生成式人工智能系统的迅猛发展，社交媒体上铺天盖地的辱骂和谎言已不再是最大的危险。最大的危险反而是对线上交流缓慢、隐蔽且具有腐蚀性的操纵——这种宣传的设计不是为了制造轰动，而是为了悄无声息地渗透到我们的日常数字讨论中来。我们已进入了国际舆论影响力行动的新时代，在这个时代里，人工智能生成的叙事会以不动声色的方式改变政治格局。

A Chinese company called GoLaxy is already undertaking such operations, according to a large cache of documents recently uncovered by the Vanderbilt University Institute of National Security, where we work. The materials show GoLaxy emerging as a leader in technologically advanced, state-aligned influence campaigns, which deploy humanlike bot networks and psychological profiling to target individuals. Its activities and claims suggest it has connections to the Chinese government.

根据我们任职的范德比尔特大学国家安全研究所最近发现的大量文件，一家名叫中科天玑(GoLaxy)的中国公司已正在进行这类活动。这些文件显示，中科天玑正在成为一场技术上先进、国家导向型的影响力行动中的领军者，通过部署类人的机器人账号网络和心理画像技术，锁定个体目标。这家公司的活动和宣称都表明它与中国政府有联系。

GoLaxy has already deployed its technology in Hong Kong and Taiwan, and the documents suggest it may be preparing to expand into the United States. A.I.-driven propaganda is no longer a hypothetical future threat. It is operational, sophisticated and already reshaping how public opinion can be manipulated on a large scale.

中科天玑已把它的技术部署到香港和台湾。文件暗示，公司也许正在为将活动扩大到美国做准备。人工智能驱动的政治宣传已不再是假想的未来威胁。这种复杂巧妙的技术已投入使用，并正在重塑大规模舆论操控的模式。

A representative of GoLaxy said the company focused on services for business intelligence and denied it had developed a bot network or psychological profiling tools targeting individuals. The company also denied being under the authority of any government agency or organization.

中科天玑的一名代表称，公司业务的重点是提供商业情报服务，并否认公司已研发了针对个体的机器人账号网络或心理画像工具。公司还否认隶属于任何政府机构或组织。

What sets GoLaxy apart is its integration of generative A.I. with enormous troves of personal data. Its systems continually mine social media platforms to build dynamic psychological profiles. Its content is customized to a person’s values, beliefs, emotional tendencies and vulnerabilities. According to the documents, A.I. personas can then engage users in what appears to be a conversation — content that feels authentic, adapts in real-time and avoids detection. The result is a highly efficient propaganda engine that’s designed to be nearly indistinguishable from legitimate online interaction, delivered instantaneously at a scale never before achieved.

中科天玑的独特之处在于它将生成式AI与海量的个体数据结合起来。公司的系统不断挖掘社交媒体平台，以构建动态的心理画像，用个体的价值观、信仰、情绪倾向以及弱点来生成定制的内容。文件显示，这些AI生成的形象能与用户用看似对话的方式互动，内容真实感强，还能实时调整，让对话者难以察觉。其结果是一个设计得与真人在线互动几乎难以区分的高效宣传引擎，把宣传以前所未有的规模即时传递给受众。

While the documents offered no specific examples of these conversations, they describe how the technology develops personalized content. By extracting user data and studying broader patterns, A.I. can build synthetic messaging designed to appeal to a wide spectrum of the public. It can adapt to a user’s tone, values, habits and interests, according to the documents. Then it can mimic real users by liking posts, leaving comments and pushing targeted content.

虽然文件没有给出这些对话的具体示例，但揭示了该技术生成个性化内容的原理。通过提取用户数据，以及对更广泛的行为模式进行分析，AI能构建吸引不同受众的合成信息。据文件描述，它能适应用户的语气、价值观、习惯和兴趣。然后，它能通过给帖子点赞、发表评论，以及推送有针对性内容等做法来模仿真实的用户。

According to the documents we uncovered, GoLaxy used its technology to minimize opposition to a 2020 national security law that cracked down on political dissent, identifying thousands of participants and thought leaders from 180,000 Hong Kong Twitter accounts. Then GoLaxy went after what it perceived as lies and misconceptions, “correcting” the sources via its army of fake profiles.

据我们发现的文件，为了把针对2020年实施的国家安全法（该法律被用于打击政治异见）的反对声音减少到最低限度，中科天玑使用其技术，从18万个香港Twitter账号中识别出几千个参与者和一件领袖的账号。然后，中科天玑针对被其认定为谎言和误解的信息，动用虚假资料建立的账号大军对这些消息源进行“纠正”。

The company struck again in the lead-up to the 2024 Taiwanese election, when China-aligned groups peddled false claims of corruption and posted deepfakes on social media. During the campaign, GoLaxy suggested ways to undermine Taiwan’s Democratic Progressive Party, which opposes China’s claims over the island. The company gathered and most likely supplied information on trends in Taiwanese political debate and recommended the deployment of bot networks to exploit political divisions between its parties. GoLaxy had already amassed an abundance of data on Taiwan to support such intrusions, according to the documents, including organizational maps of government institutions — down to their political tendencies, attitude toward China and GPS coordinates — and profiles of over 5,000 accounts belonging to Taiwanese people.

在2024年的台湾大选前夕，中科天玑再次出击。当时曾有一些亲中国的团体散布虚假的腐败指控，在社交媒体上发布深度伪造的视频。在竞选期间，中科天玑曾建议削弱台湾民进党的办法（该党反对中国对台湾的领土主张）。该公司收集并很可能提供了有关台湾政治辩论趋势的信息，并建议部署机器人账号网络，以充分利用党派政治分歧。据发现的文件，中科天玑已为支持这种干预积累了大量有关台湾的数据，包括政府机构的组织结构图——详尽到政治倾向、对中国的态度，以及GPS坐标，还有逾5000个属于台湾人账号的个人资料。

In a written statement, GoLaxy denied providing technical support for activities in Hong Kong and Taiwan.

中科天玑在一份书面声明中否认公司为香港和台湾的相关活动提供了技术支持。

So far, GoLaxy’s active deployments appear to have been confined to the Indo-Pacific. Evidence in the documents suggests that the company is positioning itself for expanded operations, including in the United States. GoLaxy has assembled data profiles of at least 117 members of the U.S. Congress and over 2,000 American political figures and thought leaders. Assuming GoLaxy continues to build American dossiers, it is possible the company will bring its operations across the Pacific. It said it has not collected data targeting U.S. officials

到目前为止，中科天玑的主动部署似乎仅限于印太地区。发现的文件中有证据暗示，该公司正为业务扩张进行布局，包括在美国的行动。中科天玑已收集了至少117名美国国会议员及2000多名美国政治人物和思想领袖的数据档案。假设中科天玑继续构建美国人的数据库，公司有可能将活动范围扩大到太平洋彼岸。中科天玑称公司没有针对美国官员收集数据。

GoLaxy operates in close alignment with China’s national security priorities, although no formal government control has been publicly confirmed. The company was founded in 2010 by a research institute at the state-controlled Chinese Academy of Sciences, and has been chaired by a deputy director from the same institute. Since then, GoLaxy has, according to the documents, worked with top-level intelligence, party and military bodies, suggesting integration with China’s political system.

中科天玑的行动与中国的国家安全优先事项密切相关，尽管没有公开资料表明公司处于政府的正式控制之下。这家公司是政府机构中国科学院的一个研究所在2010年创立的，该研究所的一名副所长一直担任公司董事长。据发现的文件，中科天玑自那时起就一直与高层情报机构、党政和军事部门合作，表明公司与中国的政治体系密不可分。

GoLaxy’s strategic alignment became clearer in 2021, when it received funding from Sugon, a Beijing-based supercomputing company flagged by the Pentagon as a Chinese military affiliate. GoLaxy’s public-facing A.I. platform coordinates with Sugon’s supercomputers and DeepSeek-R1, one of China’s leading A.I. models.

中科天玑的战略定位在2021年变得更加清晰，公司在那年获得了中科曙光的资助。总部设在北京的中科曙光是一家超级计算公司，已被五角大楼列为中国军方附属机构。中科天玑面向公众的人工智能平台与中科曙光的超级计算机以及中国领先的人工智能模型之一DeepSeek-R1有合作关系。

These connections are a reminder that influence operations are no longer a sideshow — they are becoming core instruments of statecraft. Battlefields include not only geographic territory with troops and ships but also the online platforms we use every day.

这些关系提醒人们，影响力行动不再只是一种次要活动，它们正在成为治国的核心工具。战场不只是有军队和舰艇存在的地理版图，也包括我们每天使用的网络平台。

The strategy deployed by GoLaxy and others weaponizes the openness that underpins democratic societies. Debate, transparency and pluralism — hallmarks of democratic strength — are also points of vulnerability. Technological tools like GoLaxy’s exploit these qualities. The line between surveillance and persuasion is disappearing, fast.

中科天玑等公司部署的战略是将构成民主社会基础的开放性加以武器化。辩论、透明、多元是民主力量的标志，也是其脆弱点。中科天玑等公司使用的技术工具利用的正是这些特性。监控与说服之间的界限正在迅速消失。

The danger lies in the stealth and scale of these methods, and the speed with which they are improving. A.I.-generated content can be deployed quietly across entire populations with minimal resistance. It operates continually, shaping opinion and corroding democratic institutions beneath the surface. Imagine today’s most effective social media platforms, but on a far greater scale, using a far more comprehensive model of its targets and synthetic propaganda that is even more compelling and difficult to resist.

危险在于这些方法的隐蔽性、规模以及改进的速度。人工智能生成的内容能在几乎不遇到任何阻力的情况下悄无声息地传播给整个人口。它不间断地运作，在潜移默化中塑造舆论、腐蚀民主制度。它的规模比如今能想象出来的最有效的社交媒体平台大得多，它针对目标使用的模型更全面，它合成的宣传内容更具盅惑性、更加难以抵挡。

To counter the growing threat of A.I.-driven foreign influence operations, a coordinated response is essential. Academic researchers must work urgently to map how artificial intelligence, open-source intelligence and online influence campaigns converge to serve hostile state objectives. The U.S. government must take the lead in disrupting the infrastructure behind these operations, with the Defense Department targeting foreign influence networks and the Federal Bureau of Investigation working closely with digital platforms to identify and counter false personas. The private sector needs to accelerate A.I. detection capabilities to bolster our ability to detect synthetic content. If we can’t identify it, we can’t stop it.

为消除AI驱动的外国影响力行动日益严峻的威胁，各方协调响应至关重要。从事学术研究的人需要紧急行动起来，以厘清人工智能、可公开获得的情报以及网络影响力行动是怎样汇集起来，为敌对国家的目的服务。美国政府必须在打击这些行动背后的基础设施方面发挥主导作用，让国防部打击外国影响力网络，让联邦调查局与数字平台密切合作，以确认和消除用AI伪装真人建的账号。私营部门需要加快人工智能检测能力，以增强我们发现合成内容的能力。如果我们无法识别，我们就无法阻止。

We are entering a new era of gray-zone conflict — one marked by information warfare executed at a scale, speed and degree of sophistication never seen before. If we don’t quickly figure out how to defend against this kind of A.I.-driven influence, we will be completely exposed.

我们正在进入一个灰色地带冲突的新时代，其标志是信息战正在以前所未有的规模、速度和复杂程度展开。若不能迅速找到如何防御这种AI驱动的影响力的办法，我们将彻底暴露在危险面前。