Israel-funded fake think tank floods the web to poison AI chatbots

The Hanover Institute for Public Policy presents itself as a neutral think tank studying Israel, Palestine, and antisemitism, complete with footnoted ‘data reports,’ tables of contents, and a red-white-blue American design. It is not real. A footer disclaimer reveals it was built for Israel’s Government Advertising Agency by Piro, Inc., a firm that openly markets ‘AI Story Optimization’ — content reverse-engineered to match how large language models judge credibility. The reports are structured as answers to questions a user might type into a chatbot (‘What Caused the Displacement of Palestinians in 1948?’), a format designed to surface in tools like Claude and Gemini rather than to persuade human readers.

The operation has industrial scale and clear funding. Piro has received $900,000 from Israel, routed through the French PR conglomerate Havas Media, and has pushed out more than 100 reports since it began publishing on August 6. The content appears largely machine-generated: GPTZero flagged 11 of 12 sampled articles as AI-written with high confidence. Despite claiming to cite peer-reviewed research, the pieces lean heavily on Israeli government sources like the IDF and Foreign Ministry, and most reports circle back to linking their topic to rising antisemitism. NewsGuard’s Alice Lee described the sites as near-perfect mimicry of a credible U.S. think tank, engineered around the statistics and citations that LLMs favor.

This fits a broader, well-financed Israeli push to shape AI outputs — including a separate $46.5 million contract with former Trump campaign manager Brad Parscale, whose pro-Israel sites a Drop Site investigation found had already influenced Microsoft Copilot and Google Gemini. Founder Daniel Rosenberg publicly frames the work as countering misinformation with sourced facts, but his own LinkedIn pitch is blunter: Piro spent months reverse-engineering how chatbots assemble that ‘one confident paragraph’ so brands can control what the AI says. The episode is a concrete case study in LLM poisoning as a paid influence service, and a warning about how easily authoritative-looking synthetic sources can slip into the training and retrieval pipelines behind mainstream chatbots.