Israel-backed site accused of influencing ai chatbots
A US-based website operating under the name of a non-existent think tank and reportedly funded by Israel has been producing content “with the aim of making AI chatbots promote pro-Israel arguments,” according to The Guardian.
According to an extensive analysis by the newspaper, the website, which operates under the name “Hanover Institute for Public Policy,” published 124 reports containing more than 560,000 words in just nine days, CE Report quotes Anadolu Agency.
The US-based site was built on a commercial platform that promises to help ensure that AI chatbots use its content as a source.
The website presented Israel’s position on issues such as the alleged torture of Palestinian prisoners, Israel’s alleged war crimes, and whether Israel deliberately starved Palestinians in Gaza, framing these positions as supposedly “neutral research.”
The “Hanover Institute for Public Policy” does not appear to have legal status in any jurisdiction, has no physical address, and does not list the names of its staff or authors of its reports.
New York-based media production company Piro Inc. disclosed the website to the US Department of Justice under the Foreign Agents Registration Act (FARA) as part of activities carried out on behalf of the Israeli government.
Website reportedly backed by tens of millions of dollars from Israel
The Guardian reported that the website is part of a broader campaign backed by tens of millions of dollars in funding from the Israeli government, with the activities being carried out through third parties, including the European advertising group Havas Media.
A contract between Havas Media and another US company involved in the campaign refers to deploying websites and content to generate “GPT framing results in GPT conversations.”
Of the 124 titles published by the website presenting itself as a think tank, all but one resembled questions that users might ask AI chatbots. These included: “Is anti-Zionism antisemitism?”, “Did Israel expel Palestinians from their land?”, and “Is Israel committing genocide in Gaza?”
The content was presented in a supposedly academic and well-sourced format, while the framing was designed to weaken criticism of Israel.
“As you read the texts, many things don’t make sense”
Nick Cleveland-Stout, a researcher at the US think tank Quincy Institute, said his first impression of the website was “impressive,” noting that its design and logos gave it an academic appearance. However, he said that as he read the texts, many aspects began to make little sense.
Cleveland-Stout said the content generally contained framing intended to weaken criticism of Israel, adding that almost every article ended with an implication that the subject could contribute to antisemitism.
He warned that AI chatbots could repeat such narratives without identifying their origins, making it more difficult for users to verify where the information came from.
Cleveland-Stout also said that attempts by foreign governments, companies and other actors to influence large language models have become “part of the new reality we live in.”
Photo: Chat GPT
