以色列资助的假美国智库试图利用AI进行宣传

一个打着不存在的智库名字的亲以色列信息网站在九天内发布了超过五十万字的内容，该网站基于一个商业平台，该平台承诺优化内容，以便让人工智能聊天机器人引用它。

该网站就包括折磨巴勒斯坦囚犯：https://hanoverinstitute.com/research/torture-of-prisoners-and-the-treaty-standard、以色列战争罪：https://hanoverinstitute.com/research/war-crimes-evidence-and-the-findings-issued，以及以色列是否故意让加沙巴勒斯坦人挨饿：https://hanoverinstitute.com/research/starvation-campaign-claim-and-the-records 等议题，提供了以色列的立场，所有内容都以中立研究形式呈现。

纽约媒体制作公司Piro Inc本月根据1938年的《外国代理人注册法》（Fara）向美国司法部提交了该网站：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7732，该法律要求任何为外国政府工作的个人必须正式披露其身份，作为它为以色列政府分发的材料。

该网站以汉诺威公共政策研究所的名义发布内容：https://hanoverinstitute.com，据称该机构在任何司法管辖区都不存在法律实体，没有实际地址，也没有列明的工作人员，其报告上没有署名。其自身的使用条款受“研究所建立所在州”的法律管辖，但他们拒绝说明具体州名。

该网站是由以色列政府投入数千万美元并通过包括欧洲广告集团Havas Media在内的第三方运作的更广泛努力的一部分，目的是让聊天机器人传播以色列的论点。Havas与另一家参与该运动的美国公司之间的合同明确规定，其任务是“部署网站和内容，以在GPT对话中提供GPT框架的结果”。

《卫报》联系了Piro Inc、Havas Media及以色列政府广告机构LaPam以获取评论。还通过汉诺威研究所网站上发布的联系电子邮件地址寻求评论。

这正好发生在以色列在美国舆论中的地位崩溃之际：https://www.pewresearch.org/short-reads/2026/04/07/negative-views-of-israel-netanyahu-continue-to-rise-among-americans-especially-young-people/，同时以色列官员已经开始向媒体透露，用于所谓“哈斯巴拉”海外公共外交的资金被浪费了。

根据《卫报》的一项分析，汉诺威研究所在8月6日至14日期间发布了124份报告，总计超过56万字，平均每份约4500字。仅在8月12日和13日两天，就产生了73份报告，总计近354,000字。自汉诺威研究所发布首批报告以来，再也没有发布过新的内容。

网站上124个标题中，除一个外，都是以类似用户可能在聊天机器人中输入的问题开头，包括“反犹太复国主义是反犹太主义吗？：https://hanoverinstitute.com/research/is-anti-zionism-antisemitism”、“以色列是否驱逐了巴勒斯坦人？：https://hanoverinstitute.com/research/palestinian-expulsion-and-the-1948-record”、“以色列是否在加沙实施种族灭绝？：https://hanoverinstitute.com/research/genocide-claim-and-the-convention-standard”

这些报告以准学术方式呈现，并严格注明来源，引用了包括世界银行：https://hanoverinstitute.com/research/israeli-palestinian-birth-rate-gap、联合国机构：https://hanoverinstitute.com/research/gaza-now-and-what-each-body-counts、巴勒斯坦中央统计局：https://hanoverinstitute.com/research/israeli-palestinian-population-density、大赦国际：https://hanoverinstitute.com/research/amnestys-genocide-and-apartheid-findings 以及《防止及惩治灭绝种族罪公约》的文本：https://hanoverinstitute.com/research/genocide-claim-and-the-convention-standard。

昆西研究所的研究助理尼克·克利夫兰-斯托特（Nick Cleveland-Stout）曾在该网站上报道：https://responsiblestatecraft.org/israel-influence-chatgpt/，他告诉《卫报》，他最初的印象是令人印象深刻。“它成功地伪装得非常学术——网站设计、标志，所有东西看起来都很好。但你阅读得越多，就越能意识到很多内容根本没有意义。”

例如，汉诺威研究所网站提供的信息常常被框架化，以减弱对以色列的批评。当人权组织得出结论认为以色列在实施种族灭绝或实行隔离政策时，这些报告通常会指出，没有法院对此类指控作出判决：https://hanoverinstitute.com/research/war-crimes-claims-and-the-legal-record — 根据《卫报》的统计，这种表述出现在124份报告中的20份。

伤亡和援助数据被归因于“事件的一方”：https://hanoverinstitute.com/research/what-the-record-shows-on-civilian-harm-in-gaza，这种警告在55份报告中使用，并同样适用于以色列自身的统计数据：https://hanoverinstitute.com/research/aggressor-framing-and-the-definitions，也适用于巴勒斯坦的数据。读者留下的印象是，有关加沙的情况很难确定。

一项研究——对英国、法国和德国媒体Facebook页面上的反犹太言论的2022年分析：https://www.nature.com/articles/s41599-022-01337-8 —— 在124份报告中的88份中被引用454次，包括在明显无关的主题页面中，如被占领土的土地登记：https://hanoverinstitute.com/research/land-titling-and-the-registration-record，美国媒体覆盖中的Aipac影响力：https://hanoverinstitute.com/research/aipac-and-media-coverage，加沙死亡人数：https://hanoverinstitute.com/research/gaza-death-toll-and-the-genocide-question，1948年驱逐事件：https://hanoverinstitute.com/research/palestinian-expulsion-and-the-1948-record，以及加沙封锁：https://hanoverinstitute.com/research/gaza-blockade-and-human-dignity。

克利夫兰-斯托特指出了这种引用的累积效果。“几乎每篇文章结尾都会说，好吧，仅仅谈论这个问题可能会助长反犹太主义，”他说。 “而且还会插入关于英国Facebook评论的那项研究。”

《卫报》的分析发现，该网站最初包含一种文件，这种文件告诉 AI 系统网站的内容和如何读取它（一个 llms.txt 文件），但该文件完全没有描述汉诺威研究所。它是 Res 的标准文件：https://tryres.ai，这是一个“AI 原生内容平台，帮助 B2B（企业对企业）团队被 ChatGPT、Perplexity、Claude 和 Gemini 引用”，销售行业所谓的生成引擎优化服务。

《卫报》于 8 月 13 日归档了该文件：https://web.archive.org/web/20260813104415/https://hanoverinstitute.com/llms.txt。此后，该文件已被替换为专门为机器读取者编写的版本，Res 的名字在网站上不再出现。

Res 的创始人 Hai Tran 于 7 月 22 日向美国司法部登记，作为 Piro 在以色列账户上的分包商，职位为“数字策略师”。Tran 告诉 Politico，他的角色在该登记文件中已披露。

《卫报》联系了 Res 以获取评论。

Piro 自己的网站宣传其一项名为 AI 故事优化的服务，该服务将“撰写针对 LLMs 评估可信度而优化的内容”，工作内容将“发布在您的网站或我们建立的可信第三方平台上”。

Cleveland-Stout 说，更重要的目标不是用户通过实时搜索访问网站的可能性，而是训练数据。“有两种方法，”他说。第一种是发布网站并希望聊天机器人引用它们。第二种，“更令人担忧且可能更有效的方法”，是将材料放入如 Common Crawl 这样的仓库，这些仓库为商业模型提供基础训练数据。

“当聊天机器人复述它所训练的叙述时，它甚至不会引用来源，”他说。“你将无法进行事实核查。”

较早的网络已经完成了那段旅程。由前特朗普竞选经理兼数字总监布拉德·帕斯卡尔（Brad Parscale）管理的Clock Tower X公司运营的七个网站：https://www.theguardian.com/us-news/2022/jul/12/january-6-hearings-trump-brad-parscale-texts——在Common Crawl七月索引中出现了294次，该索引为商业人工智能开发者提供训练数据。汉诺威研究所（Hanover Institute）尚未出现，其报告在那次抓取结束后才发布。它处于同一管道的起点。

汉诺威网站的每一页底部都标有外国代理披露。克利夫兰-斯托特（Cleveland-Stout）将其提供给Perplexity，他表示Perplexity告诉他，用户可以访问该网站并自行判断其可信度。“但是有多少人真的会这样做？”他说。

《卫报》在报道时对聊天机器人进行的测试显示，对汉诺威研究所产出的审查已经对聊天机器人内容产生了影响。

在被问及汉诺威研究所发布的新研究时，ChatGPT的网页聊天机器人提供了四份报告的链接：https://chatgpt.com/share/6a885dc3-8ae8-83ea-be72-be7061b9111b，但也警告称该研究所“因其资金来源及存在目的而面临重大争议”，并提供了对Quincy研究所报道的概述和链接。

根据FARA档案，派罗（Piro）通过两个工作订单签订了100万美元的合同：https://efile.fara.gov/docs/7732-Exhibit-AB-20260729-1.pdf，但这一金额与其所属的更大活动的支出相比微不足道。

派罗和Clock Tower X都是Havas Media Germany GmbH的分包商，这是一家法兰克福公司，通过Havas的西班牙分支持有，最终由阿姆斯特丹上市公司Havas NV控制。

哈瓦斯被列为五家美国公司的美国外国代理人登记中的中介：Clock Tower X：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7649，Piro：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7732，Targeted Communications Global：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7553，Bridges Partners：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7652 和 Davis Media NY：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7662。司法部的备案记录显示，它在2025年10月24日至2026年3月3日期间分六期向 Clock Tower X 支付了 1500 万美元：https://efile.fara.gov/docs/7649-Supplemental-Statement-20260518-1.pdf，并在2025年10月至12月期间向 Targeted Communications Global 支付了 9,892,335 美元：https://efile.fara.gov/docs/7553-Supplemental-Statement-20260630-2.pdf。

在首次付款七天后，即2025年10月31日，哈瓦斯自己的美国子公司：https：//efile.fara.gov/ords/fara/f？p=1381：200：：：NO：RP，200：P200_REG_NUMBER：7106 终止了其作为以色列政府代理人的注册，文件显示。在持有该注册期间，必须列举所收到的金额：其2023年10月31日前六个月的申报：https：//efile.fara.gov/docs/7106-Supplemental-Statement-20231130-3.pdf 显示以色列账户上来自其德国姊妹公司共计2,951,745美元。一旦德国实体开始直接向美国承包商支付费用，哈瓦斯便停止披露任何信息，因为外国委托人没有提交文件义务。

实际上，这意味着更广泛的运动不会立即在公开记录中显现，只能看到零散的承包商文件。

在以色列，采购记录显示，政府广告机构LaPam将合同授予Havas实体而未进行竞争性招标：2019年分别为3900万美元（3350万欧元：https://mr.gov.il/ilgstorefront/he/p/633783）、2900万美元（2500万欧元：https://mr.gov.il/ilgstorefront/he/p/641178）和1750万美元（1500万欧元：https://mr.gov.il/ilgstorefront/he/p/645070），以及2023年3月的5830万美元（5000万欧元：https://mr.gov.il/ilgstorefront/he/p/4000567498），用于计划和购买海外媒体。该合同于2024年3月15日到期。对LaPam截至2026年8月5日发布的记录以及招标登记的搜索未发现授予Havas的进一步合同——没有覆盖Havas向美国公司支付2490万美元的期间。

这笔资金并不限于用于推动聊天机器人。Targeted Communications Global的备案：https://efile.fara.gov/docs/7553-Supplemental-Statement-20260630-2.pdf 披露了一个名为"Freedom Not Terror"的社交媒体活动：https://www.freedomnotterror.org/，其30秒视频：https://www.youtube.com/watch?v=4rV-HaFjYJM，于2025年10月在YouTube上以非公开形式发布，观看次数为2810万次。其TikTok账号：https://www.tiktok.com/@freedomnotterror 有425名粉丝。该活动网站上写着“由LAPAM支付”。自2025年12月以来未发布任何内容。

《卫报》已联系目标通信全球寻求评论。

在以色列，官员们对通过操控AI来影响美国公众舆论的活动，以及公共外交整体表示怀疑。今年7月，以色列官员向Ynet：https://www.ynet.co.il/news/article/hjkka8lezx 通报该活动已失败，一位官员表示政府“花了很多钱，但情况只会变得更糟”。

以色列公共外交机构的一名前高级官员对同一家报纸表示，这“是一个与波拉德事件同样严重的可怕骗局”——指的是乔纳森·波拉德（Jonathan Pollard），这位美国分析员因在1987年为以色列间谍活动入狱：https://www.theguardian.com/world/2026/may/05/jonathan-pollard-israel-gaza-ethnic-cleansing，此事件标志着该国与华盛顿关系的低谷。

在谈到帕斯卡尔时，另一位官员对 Kan：https://www.kan.org.il/content/kan-news/politic/1076341/ 说：“他本应该改善情况。我们付给他很多钱。但他用这些钱做了什么？”外交部长吉迪恩·萨尔（Gideon Sa’ar）在上任时拨款1.816亿美元（5.45亿新谢克尔）用于国际公共外交，随后又为2026年和2027年拨款6.663亿美元（20亿新谢克尔）。

YNet 报道称，预算控制已转交给该部的常务董事埃登·巴特尔（Eden Bartel）以及部长聘请的外部顾问埃兰·沙约维奇（Eran Shayovitch）。沙约维奇在皮罗的外国代理人申报文件中被列为该公司与之接触的两名以色列官员之一。

美国舆论调查显示了这些挫败感的来源。Gallup：https://news.gallup.com/poll/702440/israelis-no-longer-ahead-americans-middle-east-sympathies.aspx 在二月发现，美国人对巴勒斯坦人的同情首次超过以色列人，仅36%的人站在以色列一边。Pew：https://www.pewresearch.org/short-reads/2026/04/07/negative-views-of-israel-netanyahu-continue-to-rise-among-americans-especially-young-people/ 记录显示，四月有60%的美国人对以色列持负面看法，比2022年增加了近20个百分点，其中年轻人变化最为明显。AP-NORC 7月7日公布的调查：https://apnorc.org/projects/democrats-are-increasingly-supportive-of-the-palestinians-while-support-for-israel-declines/ 显示，有31%的美国人表示以色列在加沙的军事行动构成种族灭绝，其中52%的民主党人持这一观点，58%的民主党人认为美国对以色列支持过度，而2024年1月这一比例为45%。

克利夫兰-斯托特怀疑任何公共外交措施是否能克服对以色列政策的反对。“人们没有反思这样一个事实，即正是这些政策导致很多美国人质疑他们对以色列的支持，”他说。“他们把一个政治问题看作是营销问题。”

他引用了已故以色列政治家约西·萨里德（Yossi Sarid）的话，他在2015年去世前一年的一次采访中开玩笑说：https://www.haaretz.com/israel-news/2025-10-15/ty-article/.premium/how-israels-hasbara-paradigm-collapsed-in-gaza/00000199-e829-da09-a1bb-faab970c0000 “Hasbara 总是所有失败者的避难所。一切都进行得很顺利——只有 Hasbara 很糟糕。如果 Hasbara 能行，那大家都能意识到理想的情况。”

然而，是否按自己的标准运作与是否对以色列有利是两个不同的问题。克利夫兰-斯托特（Cleveland-Stout）说：“Parscale 在影响聊天机器人方面取得了成功，而汉诺威研究所正并将继续在影响聊天机器人回应方面取得成功。但它们并没有改变民意调查，而这才是以色列真正关心的。”

他更广泛的担忧是这个实验所显示的内容。“无论是外国政府、公司，还是仅仅一个拥有冰淇淋店的人——LLM 的培养、投毒，或者你想怎么称呼它，都是我们生活在新的现实中的一部分。”

A pro-Israel messaging website badged with the name of a thinktank that does not exist has published more than half a million words in nine days, built on a commercial platform that promises to optimize content so that AI chatbots will cite it.

The site gives Israel’s position on subjects including the torture of Palestinian prisoners：https://hanoverinstitute.com/research/torture-of-prisoners-and-the-treaty-standard, Israeli war crimes：https://hanoverinstitute.com/research/war-crimes-evidence-and-the-findings-issued and whether Israel has deliberately starved：https://hanoverinstitute.com/research/starvation-campaign-claim-and-the-records Palestinians in Gaza, all presented as neutral research.

The New York media production company Piro Inc filed the site：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7732 with the US justice department this month under the Foreign Agents Registration Act (Fara), a 1938 law that requires anyone working for a foreign government to formally disclose it, as material it distributes for the Israeli government.

The site publishes in the name of the Hanover Institute for Public Policy：https://hanoverinstitute.com, which apparently does not exist as a legal entity in any jurisdiction, has no physical address, and carries no named staff and no bylines on any of its reports. Its own terms of use are governed by the laws of “the state in which the Institute is established”, which they decline to name.

The site is one part of a broader effort, financed by tens of millions of dollars from the Israeli government and routed through third parties including the European advertising group Havas Media, to prime chatbots to make Israel’s arguments for it. A contract between Havas and another American firm working on the campaign commits it to the “deployment of websites and content to deliver GPT framing results on GPT conversations”.

The Guardian contacted Piro Inc, Havas Media and the Israeli government advertising agency LaPam for comment. Comment was also sought via a contact email address published on the Hanover Institute website.

It comes at a point when Israel’s standing in US public opinion has collapsed：https://www.pewresearch.org/short-reads/2026/04/07/negative-views-of-israel-netanyahu-continue-to-rise-among-americans-especially-young-people/, and when Israeli officials have begun briefing media outlets that money spent on so-called “Hasbara” efforts at public diplomacy overseas has been wasted.

According to a Guardian analysis, between 6 and 14 August the Hanover Institute published 124 reports totaling more than 560,000 words, an average of about 4,500 words each. On 12 and 13 August alone it produced 73 reports, amounting to almost 354,000 words in two days. Nothing has been published since the first reports about the Hanover Institute.

All but one of the 124 headlines on the website opens with a question resembling one that a user might type into a chatbot, including “Is Anti-Zionism Antisemitism?：https://hanoverinstitute.com/research/is-anti-zionism-antisemitism”, “Did Israel Expel Palestinians From Their Land?：https://hanoverinstitute.com/research/palestinian-expulsion-and-the-1948-record”, “Is Israel Carrying Out a Genocide in Gaza?：https://hanoverinstitute.com/research/genocide-claim-and-the-convention-standard”

The reports are presented in a quasi-academic manner and are punctiliously sourced, with citations including the World Bank：https://hanoverinstitute.com/research/israeli-palestinian-birth-rate-gap, UN agencies：https://hanoverinstitute.com/research/gaza-now-and-what-each-body-counts, the Palestinian Central Bureau of Statistics：https://hanoverinstitute.com/research/israeli-palestinian-population-density, Amnesty International：https://hanoverinstitute.com/research/amnestys-genocide-and-apartheid-findings and the text of the Genocide Convention：https://hanoverinstitute.com/research/genocide-claim-and-the-convention-standard.

Nick Cleveland-Stout, a research associate at the Quincy Institute who has reported：https://responsiblestatecraft.org/israel-influence-chatgpt/ on the site, told the Guardian that his first impression was of something impressive. “It succeeds in masquerading as very academic – the website design, the logos, everything just looks good. But the more you read it, you’re able to realise that a lot of it makes no sense.”

For example, the information the Hanover Institute site presents is persistently framed to blunt criticism of Israel. Where human rights organisations have concluded that Israel is committing genocide or apartheid, the reports reliably observe that no court has entered a verdict on the claim：https://hanoverinstitute.com/research/war-crimes-claims-and-the-legal-record – a formulation that appears in 20 of the 124 reports according to the Guardian’s count.

Casualty and aid figures are attributed to “a party to the events”：https://hanoverinstitute.com/research/what-the-record-shows-on-civilian-harm-in-gaza, a caveat used in 55 reports and applied to Israel’s own counts：https://hanoverinstitute.com/research/aggressor-framing-and-the-definitions as readily as to Palestinian ones. The reader is left with the impression that little about Gaza can be established.

One study – a 2022 analysis of antisemitic comments on the Facebook pages of British, French and German media outlets：https://www.nature.com/articles/s41599-022-01337-8 – is cited 454 times across 88 of the 124 reports, including in pages on apparently unrelated topics including land registration in the occupied territories：https://hanoverinstitute.com/research/land-titling-and-the-registration-record, Aipac’s influence on US media coverage：https://hanoverinstitute.com/research/aipac-and-media-coverage, the Gaza death toll：https://hanoverinstitute.com/research/gaza-death-toll-and-the-genocide-question, the 1948 expulsions：https://hanoverinstitute.com/research/palestinian-expulsion-and-the-1948-record and the blockade of Gaza：https://hanoverinstitute.com/research/gaza-blockade-and-human-dignity.

Cleveland-Stout noted the cumulative effect of this inclusion. “Almost every article concludes with saying, well, just talking about this though could contribute to antisemitism,” he said. “And it’ll throw in the study about Facebook comments in Britain.”

The Guardian’s analysis found that initially the site carried a kind of file that tells AI systems what a site contains and how to read it (an llms.txt file), but which did not describe the Hanover Institute at all. It was the standard file of Res：https://tryres.ai, an “AI-native content platform that helps B2B [business-to-business] teams get cited by ChatGPT, Perplexity, Claude, and Gemini”, which sells what the industry calls generative engine optimization.

The Guardian archived the file：https://web.archive.org/web/20260813104415/https://hanoverinstitute.com/llms.txt on 13 August. It has since been replaced with a bespoke version written for machine readers, and Res is no longer named anywhere on the site.

Res’s founder, Hai Tran, registered with the US justice department on 22 July as a subcontractor to Piro on the Israeli account, giving his position as “digital strategist”. Tran told Politico that his role was disclosed in that filing.

The Guardian contacted Res for comment.

Piro’s own website advertises a service it calls AI Story Optimization, under which it will “author content engineered for how LLMs evaluate credibility”, with the work shipping “on your site or on trusted third-party properties we build”.

Cleveland-Stout said the more consequential target was not the possibility of users getting to the site with a live search, but the training data. “There’s kind of two ways to do it,” he said. The first is publishing sites and hoping chatbots cite them. The second, “the more concerning and potentially more effective thing”, is getting material into repositories such as Common Crawl that supply the underlying training data for commercial models.

“When it regurgitates narratives the chatbot is trained on, it won’t even cite the sources,” he said. “You won’t be able to fact-check it.”

The older network has already made that journey. The seven sites run by Clock Tower X – the firm run by former Trump campaign manager and digital director Brad Parscale：https://www.theguardian.com/us-news/2022/jul/12/january-6-hearings-trump-brad-parscale-texts – appeared 294 times in Common Crawl’s July index, the repository that supplies training data to commercial AI developers. The Hanover Institute does not appear yet, its reports having gone up after that crawl closed. It is at the start of the same pipeline.

Every page of the Hanover site carries a foreign-agent disclosure at its foot. Cleveland-Stout put that to Perplexity, which he said told him that users could visit the website and judge its trustworthiness for themselves. “But how many people are doing that?” he said.

The Guardian’s tests of chatbots at the time of reporting showed that scrutiny of the Hanover Institute’s output had already made a mark on chatbot content, however.

Asked for new research published by the Hanover Institute, ChatGPT’s webchatbot offered links to four reports：https://chatgpt.com/share/6a885dc3-8ae8-83ea-be72-be7061b9111b, but also warned that the institute was subject to “significant controversy over who funds it and why it exists”, offering a gloss of and links to Quincy Institute’s reporting.

Piro is contracted for $1m across two work orders：https://efile.fara.gov/docs/7732-Exhibit-AB-20260729-1.pdf according to Fara filings, but this sum is dwarfed by the spending on the wider campaign it belongs to.

Both Piro and Clock Tower X are subcontractors to Havas Media Germany GmbH, a Frankfurt company held through Havas’s Spanish arm and ultimately by the Amsterdam-listed Havas NV.

Havas is named as the intermediary in the US foreign-agent registrations of five American firms: Clock Tower X：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7649, Piro：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7732, Targeted Communications Global：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7553, Bridges Partners：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7652 and Davis Media NY：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7662. Justice department filings record it paying Clock Tower X $15m in six installments：https://efile.fara.gov/docs/7649-Supplemental-Statement-20260518-1.pdf between 24 October 2025 and 3 March 2026, and Targeted Communications Global $9,892,335：https://efile.fara.gov/docs/7553-Supplemental-Statement-20260630-2.pdf between October and December 2025.

Seven days after the first of those payments, on 31 October 2025, Havas’s own US subsidiary：https://efile.fara.gov/ords/fara/f?p=1381:200:::NO:RP,200:P200_REG_NUMBER:7106 terminated its registration as an agent of the Israeli government, according to filings. While it held that registration it was obliged to itemise what it received: its filing for the six months before 31 October 2023：https://efile.fara.gov/docs/7106-Supplemental-Statement-20231130-3.pdf records $2,951,745 from its German sister company on the Israeli account. Once the German entity began paying American contractors directly, Havas ceased to disclose anything, because a foreign principal has no filing obligation.

In effect, this meant that the broader campaign would not be immediately visible in the public record, only the piecemeal filings of individual contractors.

In Israel, procurement records show the government advertising agency LaPam awarding Havas entities contracts without competitive tender: $39m (€33.5m：https://mr.gov.il/ilgstorefront/he/p/633783), $29m (€25m：https://mr.gov.il/ilgstorefront/he/p/641178) and $17.5m (€15m：https://mr.gov.il/ilgstorefront/he/p/645070) across 2019, and $58.3m (€50m) in March 2023：https://mr.gov.il/ilgstorefront/he/p/4000567498 for the planning and purchase of media abroad. That contract expired on 15 March 2024. A search of LaPam’s published record to 5 August 2026, and of the tender register, found no further award to Havas – none covering the period in which Havas was paying $24.9m to American firms.

The money was not wholly confined to efforts to seed chatbots. Targeted Communications Global’s filings：https://efile.fara.gov/docs/7553-Supplemental-Statement-20260630-2.pdf disclose a social media campaign called Freedom Not Terror：https://www.freedomnotterror.org/, whose 30-second video：https://www.youtube.com/watch?v=4rV-HaFjYJM, posted as unlisted on YouTube in October 2025, has been viewed 28.1m times. Its TikTok account：https://www.tiktok.com/@freedomnotterror has 425 followers. The campaign site carries the line “Paid for by LAPAM”. It has published nothing since December 2025.

The Guardian contacted Targeted Communications Global for comment.

In Israel, officials have aired doubts about the campaign to sway US public opinion by gaming AI, and about public diplomacy in general. In July, Israeli officials briefed Ynet：https://www.ynet.co.il/news/article/hjkka8lezx that it had failed, one saying the government “paid a lot of money, but the situation only got worse”.

A former senior figure in Israel’s public diplomacy apparatus told the same paper it was “a terrible hoax on the scale of Pollard” – a reference to Jonathan Pollard, the American analyst jailed for spying for Israel in 1987：https://www.theguardian.com/world/2026/may/05/jonathan-pollard-israel-gaza-ethnic-cleansing, an incident that marked a low point in the country’s relationship with Washington.

Speaking to Kan：https://www.kan.org.il/content/kan-news/politic/1076341/ about Parscale, another official said: “He was supposed to improve the situation. We paid him a lot of money. But what did he do with it?” The foreign minister, Gideon Sa’ar, had allocated $181.6m (₪545m) to international public diplomacy on taking office, then a further $666.3m (₪2bn) for 2026 and 2027.

The YNet reporting said budget control had passed to the ministry’s director general, Eden Bartel, and to Eran Shayovitch, an external adviser brought in by the minister. Shayovitch is named on Piro’s foreign-agent filing as one of two Israeli officials with whom the firm engages.

US public opinion polling reveals the source of these frustrations. Gallup：https://news.gallup.com/poll/702440/israelis-no-longer-ahead-americans-middle-east-sympathies.aspx found in February that Americans sympathised more with Palestinians than with Israelis for the first time, with 36% siding with Israel. Pew：https://www.pewresearch.org/short-reads/2026/04/07/negative-views-of-israel-netanyahu-continue-to-rise-among-americans-especially-young-people/ recorded 60% of Americans holding an unfavourable view of Israel in April, up nearly 20 points since 2022, with the sharpest movement among the young. An AP-NORC survey published on 7 July：https://apnorc.org/projects/democrats-are-increasingly-supportive-of-the-palestinians-while-support-for-israel-declines/ found 31% of Americans say Israel’s military actions in Gaza constitute genocide, rising to 52% of Democrats, and 58% of Democrats saying the US is too supportive of Israel, against 45% in January 2024.

Cleveland-Stout doubted that any public diplomacy could overcome the opposition to Israeli policy. “There is a lack of reflection on the fact that the policies themselves are the ones that are leading to a lot of Americans questioning their support for Israel,” he said. “They view what is a political problem as a marketing one.”

He cited the late Israeli politician Yossi Sarid, who a year before his death in 2015 joked in an interview：https://www.haaretz.com/israel-news/2025-10-15/ty-article/.premium/how-israels-hasbara-paradigm-collapsed-in-gaza/00000199-e829-da09-a1bb-faab970c0000 that “Hasbara is always the refuge of all failures. Everything is going great – only the hasbara is garbage. If only the hasbara was ok, then everyone would realize the ideal situation.”

Whether the campaign works on its own terms is a separate question from whether it works for Israel, though. “Parscale is having success in influencing chatbots, and the Hanover Institute is and will continue to have success at influencing the responses that chatbots give,” Cleveland-Stout said. “But they’re not changing the polls, which is what Israel actually cares about.”

His wider concern was what the experiment demonstrates. “Whether it’s foreign governments, whether it’s companies, whether it’s even just a guy who owns an ice cream shop – LLM grooming, poisoning, whatever you want to call it, is part of the new reality that we live in.”