Israel-funded fake thinktank used 'AI optimization' to spread propaganda

The Israeli government funded a sophisticated influence operation that created a fake U.S.-based thinktank to manipulate the information AI chatbots provide to users. The campaign sought to frame government positions on war crimes and the treatment of Palestinian prisoners as neutral academic research.

Operating under the name of the "Hanover Institute," a non-existent organization, the website published more than 500,000 words of content within a single nine-day window. According to reports from The Guardian, the operation was distributed by Piro Inc. on behalf of Havas Media Germany GmbH for the Israel Government Advertising Agency, known as LaPam. To maximize its reach, the site was built on a commercial platform specifically designed to optimize content so that AI chatbots are more likely to cite it as a source.

The Strategy of AI Gaming

This operation targeted some of the most contentious aspects of the current conflict, including the starvation of Palestinians in Gaza and the torture of prisoners. By presenting these narratives through the lens of a supposed American research institute, the campaign attempted to influence the retrieval-augmented generation (RAG) processes and training data that power modern AI models. This approach allows state actors to bypass traditional social media filters and embed propaganda directly into the "factual" summaries generated by AI.

A Shift in Influence Operations

This represents a critical evolution in state-sponsored disinformation, moving from the era of social media bot farms to what is known as "AI Engine Optimization" (AEO). While traditional propaganda relies on viral sharing, AEO targets the underlying data sources that AI models trust. By lending propaganda an air of academic authority through a fake thinktank, state actors can automate the dissemination of their narratives at scale, making the misinformation appear as a consensus of expert research.

Future Implications

As more users rely on AI chatbots for rapid information retrieval rather than traditional search engines, the vulnerability of these models to targeted data poisoning increases. The Hanover Institute case highlights a growing arms race between AI developers and state actors seeking to weaponize the perceived neutrality of machine learning. It remains to be seen how AI companies will update their citation filters to detect and exclude content from synthetic organizations designed specifically to game their systems.