A fake think tank created on behalf of the Israeli government has published more than 100 polished research reports in just over a week, in what analysts describe as a deliberate campaign to manipulate the answers that AI chatbots give users about Israel and Palestine, according to an investigation published Monday by Responsible Statecraft.
The organization, called the Hanover Institute for Public Policy, presents itself as a new research body focused on Israel/Palestine. But its reports carry no bylines, and a disclaimer at the bottom of its website acknowledges that it was created on behalf of the Israeli Government Advertising Agency by Piro, Inc., a firm co-founded by Daniel Rosenberg, a film producer credited on Spike Lee's "Inside Man." The story was picked up by outlets including Arab News and The Cradle, and is the latest sign that influence operations are shifting from social media feeds to AI models. For more on how manipulation is evolving in the AI era, follow our breaking AI news coverage. For more context on this story, see our ongoing AI trends.
Reports Engineered for How LLMs Judge Credibility
What makes the Hanover Institute unusual is how precisely its output appears tailored to the way large language models evaluate sources. Its "data reports" come with footnotes, tables of contents, and a measured, neutral tone — the kind of formatting that helps documents appeal to chatbots such as Claude or Gemini, Responsible Statecraft reported.
Many of the titles are phrased as the exact questions a user might type into a chatbot: "What Caused the Displacement of Palestinians in 1948?" "Which Humanitarian Organizations Have Documented Israeli War Crimes?" "What is the Current Situation in the Gaza Strip?" and "Is the IDF the World's Most Moral Army?" — the last of which cites a 2022 poll finding that 47% of Israeli Jews believed that description.
Piro's own website is blunt about the method. The company says it authors "content engineered for how LLMs evaluate credibility," a service it markets as "AI Story Optimization." Researchers have a blunter name for the practice: "LLM poisoning."
A 'Perfect Mimicry' of a Credible Institution
Alice Lee, an analyst at NewsGuard, the disinformation tracking company, told Responsible Statecraft that the site's reports appear designed to reach a U.S. audience curious about the conflict, whether through search engines or AI assistants.
"LLMs favor concrete statistics and data, as well as strong citations and sources, which these articles all have," Lee said. "It's a perfect mimicry of a typical credible American think tank, right down to the generic name, the site layout, and the red-white-blue color scheme."
The institute has also published what it claims are original findings — for example, a study asserting that 19 of the 36 most-watched Israel-Gaza explainer videos contain contested claims, with most of those contested claims aligning with the Palestinian narrative. Other reports push back against established documentation, including one that casts doubt on UNICEF's assertion that 90% of water and institutional infrastructure in Gaza has been damaged or destroyed.
Notably, the output does not always track the Israeli government line. One report describes the theory that foreign funding of universities explains antisemitic incidents as "weak and full of exceptions" — even though Prime Minister Benjamin Netanyahu has promoted that theory, telling Breitbart that foreign actors have spent "billions" on American campuses. Many reports conclude by linking the topic at hand to rising antisemitism, frequently citing the same studies.
Follow the Money: $900,000 Through a French PR Conglomerate
According to Responsible Statecraft, Piro, Inc. has received $900,000 from the Israeli government for its work, subcontracted through Havas Media, a French public relations conglomerate — an arrangement consistent with how other Israeli contractors have been paid for overseas influence work.
The Hanover Institute began publishing on August 6. In the roughly ten days since, it has churned out more than 100 reports, a pace that no genuine policy shop of comparable size sustains.
Why Chatbot Manipulation Is Different
The case underscores a structural weakness in how AI assistants assemble answers: they treat well-cited, statistics-heavy documents as credible regardless of who produced them or why. Traditional disinformation targeted human readers on social platforms, where emotional punch mattered most. So-called LLM poisoning instead targets the retrieval and ranking layers that decide which sources a model quotes — and a single persuasive document can then echo into millions of answers.
For chatbot developers, the Hanover Institute is a stress test of source vetting. NewsGuard's assessment suggests the mimicry is close enough to fool automated credibility signals. For readers, the lesson is simpler: the confident, well-footnoted answer a chatbot gives you about a contested conflict may trace back to an organization that was built, branded, and funded to produce exactly that answer.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
