Israel's FARA-funded influence push now feeds ChatGPT directly
Governments have learned that the way to win an argument in 2026 is to be what the chatbot cites. Israel's latest operation shows how far that playbook has gone.
Israel has stood up a new, FARA-registered influence operation built specifically to shape what ChatGPT and other LLMs say about Gaza and the IDF — and it is already getting cited. Politico reported this week that Justice Department filings from the past seven days reveal the Hanover Institute for Public Policy, a website publishing dozens of authorless "data reports," is the work of a $100,000 campaign run by boutique ad agency Piro, Inc. on behalf of Havas Media and Israel's Government Advertising Agency (LaPam). The reports are framed as questions — is the IDF the world's most moral army, is there a policy of starvation in Gaza — a format that reads like it was engineered for retrieval by language models rather than for human readers. In neutral tests run by Politico, both ChatGPT and Perplexity cited the institute's material in their answers.
It is a second front in the same campaign, not a first: Havas previously hired Brad Parscale's Clock Tower X for a $46.5 million influence push whose FARA filings explicitly mention "GPT framing" — publishing searchable, pro-Israel content so models surface it in responses. The Hanover site carries a disclosure that it is distributed by Piro on behalf of Havas Media Germany on behalf of LaPam, and it appears to have been built with help from Res, a Seattle startup that sells "AI recommendation" services to Nike and Intel. An identical version of the site hosted on Res's own domain went dark after Politico reached out for comment.
Why it matters: a chatbot answer now carries the authority of a headline, but with none of the sourcing. When a government-funded source gets cited without its funding being visible, readers absorb advocacy as fact. The campaign is fully legal — registered under FARA, disclosed at the bottom of the page — which is exactly why it is hard to stop: the disclosure is there, and nobody scrolling a chat answer will ever see it. The interesting test is whether OpenAI, Google and Perplexity treat government-funded content farms as a transparency problem worth solving, or as a cost of doing business.
What to watch: whether the major labs respond to the Res connection with stronger provenance signals in answers, and whether other governments copy the "institute" template.
If a chatbot cites a government-funded source without telling you, is that a disclosure failure or just PR? Tell us in the comments.
