When Russia interfered in the 2016 U.S. presidential election, spreading divisive and inflammatory posts online to stoke outrage, its posts were brash and riddled with spelling errors and strange syntax. They were designed to get attention by any means necessary.

在干预2016年美国总统大选时，俄罗斯在网上传播制造不合、煽动情绪的帖子，以期挑起怒火，那些帖子显得草率，充满了拼写错误和奇怪的语法。它们的意图是尽一切必要手段引起注意。

“Hillary is a Satan,” one Russian-made Facebook post read.

“希拉里是撒旦，”一个俄罗斯创建的Facebook帖子说。

Now, eight years later, foreign interference in American elections has become far more sophisticated, and far more difficult to track.

八年后的今天，外国对美国选举的干预变得更精密、更隐蔽。

Disinformation from abroad — particularly from Russia, China and Iran — has matured into a consistent and pernicious threat, as the countries test, iterate and deploy increasingly nuanced tactics, according to U.S. intelligence and defense officials, tech companies and academic researchers. The ability to sway even a small pocket of Americans could have outsize consequences for the presidential election, which polls generally consider a neck-and-neck race.

据美国情报和国防官员、科技公司和学术研究人员称，来自境外的不实信息——主要来自俄罗斯、中国和伊朗——已经成熟，变成了一种稳固而致命的威胁，这些国家在测试、迭代和部属越来越精细的战术。即便只是改变一小群美国人的立场，都有可能对总统选举的结果产生巨大的影响，民调显示这场选举处于势均力敌的态势。

Russia, according to American intelligence assessments, aims to bolster the candidacy of former President Donald J. Trump, while Iran favors his opponent, Vice President Kamala Harris. China appears to have no preferred outcome.

据美国的情报评估，俄罗斯希望推动前总统唐纳德·J·特朗普当选，而伊朗偏向他的对手贺锦丽副总统。中国似乎没有什么倾向。

But the broad goal of these efforts has not changed: to sow discord and chaos in hopes of discrediting American democracy in the eyes of the world. The campaigns, though, have evolved, adapting to a changing media landscape and the proliferation of new tools that make it easy to fool credulous audiences.

但它们的行动的总体目标没有改变，那就是播下不和与混乱的种子，从而让美国民主在世界面前失去信誉。不过行动本身已经有所变化，在适应一个不断改变的媒体环境，以及新工具的普及，这些工具让它们可以轻易地愚弄那些轻信的受众。

Here are the ways that foreign disinformation has evolved:

以下是境外不实信息所经历的演变：

Now, disinformation is basically everywhere.

不实信息如今基本上已经无处不在。

Russia was the primary architect of American election-related disinformation in 2016, and its posts ran largely on Facebook.

俄罗斯是2016年美国选举相关的不实信息的主要构建者，其帖文基本上发表在Facebook上。

Now, Iran and China are engaging in similar efforts to influence American politics, and all three are scattering their efforts across dozens of platforms, from small forums where Americans chat about local weather to messaging groups united by shared interests. The countries are taking cues from one another, although there is debate over whether they have directly cooperated on strategies.

现在伊朗和中国也在进行类似的美国政治影响行动，这三个国家都覆盖了十几个平台，其中既有美国人聊当地天气的小论坛，也有因为共同的兴趣而结成的聊天群组。这些国家在相互借鉴，不过尚不确定它们是否有直接的战略合作。

There are hordes of Russian accounts on Telegram seeding divisive, sometimes vitriolic videos, memes and articles about the presidential election. There are at least hundreds more from China that mimicked students to inflame the tensions on American campuses this summer over the war in Gaza. Both countries also have accounts on Gab, a less prominent social media platform favored by the far right, where they have worked to promote conspiracy theories.

在Telegram上有大量俄罗斯账号在传播制造不和、有时极为尖刻的视频、米姆和文章，都与总统大选有关。至少还有数百个来自中国的账号伪装成学生，在今夏的美国校园内煽动因加沙战争而起的分歧。两个国家在Gab上也都有账号，这是个名气略小的社交媒体平台，受极右翼人士青睐，他们在这里致力于宣传阴谋论。

伊朗、俄罗斯和中国都在进行类似的美国政治影响行动，覆盖了在十多个平台。

Russian operatives have also tried to support Mr. Trump on Reddit and forum boards favored by the far right, targeting voters in six swing states along with Hispanic Americans, video gamers and others identified by Russia as potential Trump sympathizers, according to internal documents disclosed in September by the Department of Justice.

据司法部在9月公开的内部文件，俄罗斯行动人员还试图在Reddit和极右翼偏爱的论坛上支持特朗普，针对六个摇摆州的选民、西语裔美国人、游戏玩家以及其他被俄罗斯认为可能支持特朗普的人。

One campaign linked to China’s state influence operation, known as Spamouflage, operated accounts using a name, Harlan, to create the impression that the source of the conservative-leaning content was an American, on four platforms: YouTube, X, Instagram and TikTok.

与中国政府影响行动有关联的“垃圾伪装”(Spamouflage)，在四个平台上——YouTube、X、Instagram和TikTok——使用了一个名为Harlan的账号，以求制造这些有保守派倾向的内容来自一个美国人的假象。

The content is far more targeted.

内容的针对性大大增加。

The new disinformation being peddled by foreign nations aims not just at swing states, but also at specific districts within them, and at particular ethnic and religious groups within those districts. The more targeted the disinformation is, the more likely it is to take hold, according to researchers and academics who have studied the new influence campaigns.

这些外国传播的新型不实信息并不只是针对摇摆州，还针对州内的具体选区，以及区内的具体民族和宗教团体。据研究了这种新型影响行动的研究人员和学者说，不实信息越有针对性，就越有可能产生持久的效果。

“When disinformation is custom-built for a specific audience by preying on their interests or opinions, it becomes more effective,” said Melanie Smith, the research director for the Institute for Strategic Dialogue, a research organization based in London. “In previous elections, we were trying to determine what the big false narrative was going to be. This time, it is subtle polarized messaging that strokes the tension.”

“当不实信息是为具体受众度身定制，以影响他们的兴趣或观点，那就更有效了，”伦敦研究机构战略对话研究所的研究总监梅兰妮·史密斯说。“在以往的选举中，我们会试图确定最大的那个虚假叙事是什么。这一次，撩拨情绪的是一些隐晦的两极化讯息。”

Iran in particular has spent its resources setting up covert disinformation efforts to draw in niche groups. A website titled “Not Our War,” which aimed to draw in American military veterans, interspersed articles about the lack of support for active-duty soldiers with virulently anti-American views and conspiracy theories.

伊朗尤其将大量资源用于开展隐秘的不实信息行动，以吸引一些具体的人群。一个名为“不是我们的战争”(Not Our War)的网站旨在吸引美国退伍军人，在有关现役军人得不到支持的文章中夹杂恶毒的反美观点和阴谋论。

Other sites included “Afro Majority,” which created content aimed at Black Americans, and “Savannah Time,” which sought to sway conservative voters in the swing state of Georgia. In Michigan, another swing state, Iran created an online outlet called “Westland Sun” to cater to Arab Americans in suburban Detroit.

其他的网站包括“非洲多数派”(Afro Majority)，制作面向美国黑人的内容，以及“萨凡纳时间”(Savannah Time)，意在改变摇摆州乔治亚的保守派选民的立场。在另一个摇摆州密西根，伊朗创建了一个名叫“韦斯特兰太阳”(Westland Sun)的网媒，面相底特律郊区的阿拉伯裔美国人。

“That Iran would target Arab and Muslim populations in Michigan shows that Iran has a nuanced understanding of the political situation in America and is deftly maneuvering to appeal to a key demographic to influence the election in a targeted fashion,” said Max Lesser, a senior analyst at the Foundation for Defense of Democracies.

“伊朗会去针对密西根的阿拉伯和穆斯林人口，这表明它对美国的政治局势有着深切的理解，正在进行高明的运作，用有针对性的方式说服一个关键人群，以求影响选举，”保卫民主基金会高级分析师麦克斯·莱瑟尔说。

China and Russia have followed a similar pattern. On X this year, Chinese state media spread false narratives in Spanish about the Supreme Court, which Spanish-speaking users on Facebook and YouTube then circulated further, according to Logically, an organization that monitors disinformation online.

中国和俄罗斯也采用了类似的模式。据监控网络不实信息的Logically称，今年在X上，中国官方媒体用西班牙语散布了有关美国最高法院的虚假叙事，讲西班牙语的用户在Facebook和YouTube上做了进一步的传播。

Experts on Chinese disinformation said that inauthentic social media accounts linked to Beijing had become more convincing and engaging and that they now included first-person references to being an American or a military veteran. In recent weeks, according to a report from Microsoft’s Threat Analysis Center, inauthentic accounts linked to China’s Spamouflage targeted House and Senate Republicans seeking re-election in Alabama, Tennessee and Texas.

研究中国不实信息的专家表示，和北京有关联的伪造社交媒体账号已经变得更具说服力，有更多的互动，并且现在加入了身为美国人或退伍军人的第一人称指示。据微软威胁分析中心的一份报告，近几周，与中国的“垃圾伪装”关联的伪造账号针对了在阿拉巴马、田纳西和得克萨斯等州寻求连任的众议院和参议院共和党人。

Artificial intelligence is propelling this evolution.

人工智能在推动这种演变。

Recent advances in artificial intelligence have boosted disinformation capabilities beyond what was possible in previous elections, allowing state agents to create and distribute their campaigns with more finesse and efficiency.

近年人工智能的进步已经将不实信息的效力推向以往选举无法想象的程度，让政府代理人可以更精巧而高效地制作和发布。

OpenAI, whose ChatGPT tool popularized the technology, reported this month that it had disrupted more than 20 foreign operations that had used the company’s products between June and September. They included efforts by Russia, China, Iran and other countries to create and fill websites and to spread propaganda or disinformation on social media — and even to analyze and reply to specific posts. (The New York Times sued OpenAI and Microsoft last year for copyright infringement of news content; both companies have denied the claims.)

将人工智能技术推向大众的ChatGPT工具的开发者OpenAI本月发布报告称，在6月到9月间，公司瓦解了超过20项外国利用其产品进行的行动。其中包括俄罗斯、中国、伊朗以及其它国家，它们创建网站并填入内容，在社交媒体上开展宣传或不实信息传播——甚至会对具体帖文进行分析和回复。（纽约时报去年因新闻内容版权侵犯问题起诉了OpenAI和微软，两家公司都否认这一诉讼主张。）

“A.I. capabilities are being used to exacerbate the threats that we expected and the threats that we’re seeing,” Jen Easterly, the director of the Cybersecurity and Infrastructure Security Agency, said in an interview. “They’re essentially lowering the bar for a foreign actor to conduct more sophisticated influence campaigns.”

“AI的能力被用于加剧我们预料到和正在看到的威胁，”网络安全和基础设施安全局局长珍·伊斯特里在采访中说。“基本上就是在降低外国行为者开展更复杂的影响行动的门槛。”

The utility of commercially available A.I. tools can be seen in the efforts of John Mark Dougan, a former deputy sheriff in Florida who now lives in Russia after fleeing criminal charges in the United States.

透过约翰·马克·杜根的所作所为可以看到商业人工智能工具的实用性。杜根曾是佛罗里达州的一位副警长，在受到刑事指控后，他逃离美国，现居住在俄罗斯。

Working from an apartment in Moscow, he has created scores of websites posing as American news outlets and used them to publish disinformation, effectively doing by himself the work that, eight years ago, would have involved an army of bots. Mr. Dougan’s sites have circulated several disparaging claims about Ms. Harris and her running mate, Gov. Tim Walz of Minnesota, according to NewsGuard, a company that has tracked them in detail.

他在莫斯科的一间公寓里工作，创建了数十个冒充美国新闻媒体的网站，并利用这些网站发布虚假信息，基本上他一个人就完成了八年前需要一大批机器水军才能完成的工作。据详细追踪相关活动的NewsGuard公司称，杜根的网站散布了多条对贺锦丽及其竞选搭档、明尼苏达州州长蒂姆·沃尔兹的贬损言论。

美国司法部表示，俄罗斯行动人员一直在Reddit和极右翼青睐的论坛上支持前总统特朗普

China, too, has deployed an increasingly advanced tool kit that includes A.I.-manipulated audio files, damaging memes and fabricated voter polls in campaigns around the world. This year, a deepfake video of a Republican congressman from Virginia circulated on TikTok, accompanied by a Chinese caption falsely claiming that the politician was soliciting votes for a critic of Beijing who sought (and later won) the Taiwanese presidency.

中国也在全世界各地的竞选活动中部署了一套越来越先进的工具，其中包括经过人工智能篡改的音频文件、具有破坏性的米姆和伪造的民意调查。今年，一段关于弗吉尼亚州共和党众议员的深度伪造视频在TikTok上流传，并附有中文字幕，该视频谎称这位政治人士正在为一位批评北京的人士拉票，后者曾竞选（并最终赢得）台湾总统大选。

It’s becoming much harder to identify disinformation.

对虚假信息的识别变得愈发困难。

All three countries are also becoming better at covering their tracks.

这三个国家也越来越善于掩盖自己的踪迹。

Last month, Russia was caught obscuring its attempts to influence Americans by secretly backing a group of conservative American commentators employed through Tenet Media, a digital platform created in Tennessee in 2023.

上个月，俄罗斯被发现试图通过暗中支持一群通过Tenet Media雇用的美国保守派评论员，掩盖其影响美国人的企图。Tenet Media是2023年创建于田纳西州的数字平台。

The company served as a seemingly legitimate facade for publishing scores of videos with pointed political commentary as well as conspiracy theories about election fraud, Covid-19, immigrants and Russia’s war with Ukraine. Even the influencers who were covertly paid for their appearances on Tenet said they did not know the money came from Russia.

该公司充当了看似合法的幌子，发布了数十段带有尖锐政治评论观点的视频，以及关于选举舞弊、新冠疫情、移民和俄乌战争的阴谋论。即使是那些因在Tenet上露面而秘密获得报酬的博主也表示，他们不知道这些钱来自俄罗斯。

In an echo of Russia’s scheme, Chinese operatives have been cultivating a network of foreign influencers to help spread its narratives, creating a group described as “foreign mouths,” “foreign pens” and “foreign brains,” according to a report last fall by the Australian Strategic Policy Institute.

根据澳大利亚战略政策研究所去年秋季的一份报告，与俄罗斯的阴谋类似，中国特工一直在培养一个外国博主网络来帮助传播其言论，创造了一个被描述为“外国嘴”、“外国笔”和“外国脑”的群体。

The new tactics have made it harder for government agencies and tech companies to find and remove the influence campaigns — all while emboldening other hostile states, said Graham Brookie, the senior director at the Atlantic Council’s Digital Forensic Research Lab.

大西洋理事会数字取证研究实验室高级主管格雷厄姆·布鲁克表示，这些新策略使政府机构和科技公司更难发现和消除相关影响力活动，同时让其它敌对国家变得更为大胆。

“Where there is more malign foreign influence activity, it creates more surface area, more permission for other bad actors to jump into that space,” he said. “If all of them are doing it, then the cost for exposure is not as high.”

“哪里有更多恶性外国影响力活动，哪里就会创造更大的表面空间，让其他不良行为者得到更多进入该空间的许可，”他说。“如果他们都这样做，那曝光的代价就不会那么大了。” 

Technology companies aren’t doing as much to stop disinformation.

科技公司并没有采取太多措施阻止虚假信息。

The foreign disinformation has exploded as tech giants have all but given up their efforts to combat disinformation. The largest companies, including Meta, Google, OpenAI and Microsoft, have scaled back their attempts to label and remove disinformation since the last presidential elections. Others have no teams in place at all.

科技巨头几乎放弃了打击虚假信息的努力，外国虚假信息则呈爆炸式增长。自上届总统大选以来，包括Meta、谷歌、OpenAI和微软在内的大型公司已经减少了标记和删除虚假信息的尝试。其它公司则根本没有组建任何团队。

The lack of cohesive policy among the tech companies has made it impossible to form a united front against foreign disinformation, security officials and executives at tech companies said.

科技公司安全官员和高管表示，由于科技公司之间缺乏统一的政策，因此无法形成统一战线来对抗外国虚假信息。

“These alternative platforms don’t have the same degree of content moderation and robust trust and safety practices that would potentially mitigate these campaigns,” said Mr. Lesser of the Foundation for Defense of Democracies.

“这些替代平台没有同等程度的内容审核以及强有力的信任和安全实践，而相关措施可能会减轻这些活动的影响，”保卫民主基金会的莱瑟尔说。

He added that even larger platforms such as X, Facebook and Instagram were trapped in an eternal game of Whac-a-Mole as foreign state operatives quickly rebuilt influence campaigns that had been removed. Alethea, a company that tracks online threats, recently discovered that an Iranian disinformation campaign that used accounts named after hoopoes, the colorful bird, recently resurfaced on X despite having been banned twice before.

他还说，即使是X、Facebook和Instagram等大型平台也陷入了一场永无止境的打地鼠游戏，因为一旦相关影响力活动遭到移除，外国政府特工就会迅速将其重新打造起来。追踪在线威胁的公司Alethea最近发现，一个使用以戴胜鸟（一种色彩鲜艳的鸟类）命名的账户的伊朗虚假信息活动最近又出现在X上，尽管此前该账户已两次被封禁。