Israel creates fake think tank to dupe AI chatbots
Archived — this story has rotated out of today’s deck. It is kept here in full.
The gist
Israel created fake think tank Hanover Institute to influence AI chatbots. The campaign exploits AI's inability to distinguish propaganda from legitimate research.
Background
The Israeli government has been accused of creating a fake think tank, the Hanover Institute, to influence AI chatbots. The operation, part of a $46.5 million contract with Brad Parscale, aims to poison AI datasets with pro-Israel narratives about the Gaza conflict. The tactic, known as 'data void filling' or 'LLM poisoning,' exploits AI's tendency to prioritize formally structured, heavily linked content.
How it unfolded
- recentlyDrop Site investigation found that chatbots like Microsoft Copilot and Google Gemini had been trained on data from pro-Israel websites created by Brad Parscale.
- Mon, 17 Aug 2026Responsible Statecraft reported that the Hanover Institute is not a real think tank, created by Piro, Inc on behalf of the Israeli Government Advertising Agency.
- Mon, 17 Aug 2026Streamline published an exposé detailing the $46.5 million campaign to poison AI chatbots with Gaza war propaganda.
Who’s saying what
- Analysts
- Experts warn that this campaign erodes global trust in AI and highlights vulnerabilities in AI search engines.
- Public
- Critics argue that the operation is a deceptive attempt to manipulate public opinion and undermine AI integrity.
Still unverified
The effectiveness of the campaign in actually influencing AI outputs is not fully verified; the claim that chatbots were 'successfully trained' on the data is based on a Drop Site investigation.