Israel-linked fake think tank used to influence AI chatbots

The Hanover Institute for Public Policy, a non-existent entity, has been identified as a front for a pro-Israel messaging campaign designed to influence artificial intelligence chatbots. The organization produces large volumes of reports on sensitive topics, such as the situation in Gaza and Israeli military conduct, presented in a neutral, academic tone to increase their credibility with Large Language Models (LLMs).

Investigations reveal the site was created by Piro, Inc., a media production company, on behalf of the Israeli Government Advertising Agency. Piro, Inc. has filed the material with the U.S. Department of Justice under the Foreign Agents Registration Act (FARA). The campaign, which involves significant funding from the Israeli government routed through third parties like Havas Media, utilizes a service described as ‘AI Story Optimization’ to ensure chatbots like Claude or Gemini cite the institute’s content.

This practice, sometimes referred to as ‘LLM poisoning,’ involves deploying engineered content to achieve specific ‘GPT framing results’ during AI conversations. The Hanover Institute lacks a physical address, named staff, or legal existence in any jurisdiction, and its reports typically lack bylines.