How Israel has created a fake 'think tank' to manipulate AI responses

Israeli officials have created a fake think tank to manipulate responses generated by artificial intelligence, according to an investigation published by The Guardian. A website operated by pro-Israel advocacy groups, posing as a non-existent research institute, has posted over half a million words in just nine days. The site uses a commercial platform designed to optimize content so that AI chatbots will cite its materials. The website presents Israeli positions on issues such as the torture of Palestinian prisoners, alleged war crimes committed by Israel, and whether Israel deliberately used starvation against the population of Gaza. It frames these views as if they were findings from a neutral, expert-led study. The site claims to represent the Hanover Institute for Public Policy, which does not appear to exist as a legal entity in any jurisdiction, lacks a physical address, and fails to identify its staff or authors. In August, the New York-based communications agency Piro Inc. registered the website with the U.S. Department of Justice under the Foreign Agents Registration Act (FARA). This law requires individuals working for foreign governments to formally declare their relationship. Piro identified the site as distributing material on behalf of the Israeli government. The website was exposed on August 14. Since then, the institute added a page titled “Funding,” stating that a previous version claimed it operated without external funding, a claim that was true when the section was written. The site is part of a broader initiative funded with tens of millions of dollars by the Israeli government and channeled through third parties, including the European advertising group Havas Media. A contract between Havas and another U.S.-based firm involved in the campaign outlines efforts to launch websites and content aimed at shaping conversations with GPT models to promote arguments favorable to Israel. The Guardian contacted Piro Inc., Havas Media, and the Israeli government’s publicity agency, LaPam, seeking comment. The outlet also requested a response via an email address listed on the Hanover Institute’s website. This activity occurs amid a decline in Israel’s public image in the United States and growing criticism from Israeli officials who argue that funds allocated to the so-called “Hasbara”, Israel’s public diplomacy abroad, have been wasted. The site’s content includes more than 560,000 words in 124 reports between August 6 and 14. On August 12 and 13 alone, it released 73 reports totaling nearly 354,000 words. After initial exposure, publication ceased, though all but one of the 124 headlines began with questions akin to those users might ask in a chatbot, such as “Is anti-Zionism antisemitism?” or “Did Israel expel Palestinians from their land?” The reports mimic academic studies, citing detailed references to the World Bank, UN agencies, the Palestine Central Statistics Office, Amnesty International, and the text of the Genocide Convention. In statements to The Guardian, Nick Cl, a spokesperson for the Israeli Ministry of Foreign Affairs, stated that the ministry had no knowledge of the operation and emphasized that all official communication adhered to strict guidelines. The strategy reflects a broader effort to influence AI systems by feeding them information that aligns with specific narratives. By creating a false think tank, proponents aim to make their perspectives appear credible within algorithmic frameworks. Such tactics raise concerns about the integrity of information disseminated through AI-driven platforms and the potential for manipulation in political discourse. As investigations continue, the implications for transparency and accountability in digital spaces remain unclear.