Forensic Research Archive of Manipulation and Exploitation
Illustration: a contemporary investigator's evidence wall of printed social-media posts, network diagrams and a pinned world map, in the same dossier style as the rest of the archive.

Watch briefs · September 10, 2026

Israeli government advertising agency funds synthetic US think-tank Hanover Institute to seed AI retrieval and chatbot answers via high-volume question-formatted reports

Illustration, not an archive object

Filed
September 10, 2026, 21:00 UTC
Actor
Israel Government Advertising Agency (LaPam) via Havas Media Germany GmbH and US registrant Piro Inc (FARA 7732); synthetic surface Hanover Institute for Public Policy
Target
US and English-language AI chatbots, search retrieval systems, and downstream human users of those systems on Israel-Palestine questions
Source
Unidentified Source blurred: contractors, cutouts, buried disclosure
Confidence
documented Primary documents, court records, official disclosures
Techniques
Narrative launderingCorpus poisoningSynthetic mediaDisclosure theaterManufactured consensus
Where
United States, Israel, Gaza
Campaign
Hanover Institute for Public Policy (LaPam / Piro / Havas)
Channels
  • Hanover Institute website (question-formatted data reports optimized for LLM citation)
  • FARA-registered distribution by Piro Inc under Havas/LaPam work orders
  • llms.txt and generative-engine-optimization packaging
  • page-foot foreign-agent disclosure that does not travel into chatbot answers
  • high-volume unsigned academic-style output for indexing and training corpora

Summary

FARA filings and multi-outlet investigations in August 2026 establish that Piro Inc, under contract to Havas Media Germany acting for Israel’s LaPam (Government Advertising Agency), stood up the Hanover Institute for Public Policy website. The site produced 124 unsigned, AI-assisted academic-style reports totaling over 560,000 words in nine days in August, structured as chatbot-style questions on Gaza, antisemitism, war crimes and related topics. Disclosure of foreign-agent status appears on pages and in FARA but does not travel with cited content. Politico tests showed ChatGPT and Perplexity citing the material. The design goal is generative-engine optimization and corpus seeding rather than conventional human persuasion.

Analysis

Primary evidence is the public FARA registration 7732 for Piro Inc, the accompanying work-order and invoice filings that name Havas Media Germany acting for LaPam, the Hanover site’s own funding and about pages that now acknowledge the arrangement, and contemporaneous reporting by Politico (14 Aug), the Guardian (26 Aug), 404 Media (25 Aug) and others. Volume metrics (124 reports, >560k words in nine days, peaks of dozens of reports per day) are measured from the live site archive. Politico and subsequent tests confirmed chatbot citation of the material on neutral queries about Gaza, anti-Zionism and antisemitism. AI-detection samples and residual platform artifacts further support synthetic production.

Mechanism: narrative-laundering is achieved by packaging state-aligned framing inside a think-tank costume that presents as independent US research. Corpus-poisoning is the explicit design goal—content is question-formatted, volume-scaled and packaged with machine-readable signals so that search and LLM retrieval systems treat it as citable primary or secondary literature. Synthetic-media supplies the production rate no conventional research staff could match. Disclosure-theater is the page-foot and FARA filing that satisfy the letter of US law while remaining invisible to the chatbot user who receives an answer citing “the Hanover Institute.” Manufactured-consensus follows if multiple retrieval surfaces or models independently surface the same framed reports.

This is propaganda and an influence operation because the institutional form is performed (no legal entity, no named scholars, no physical address, no bylines), the funding chain is state advertising, and the load-bearing objective is to alter the source layer that machines and later humans treat as independent evidence. Technique scoring is actor-agnostic. Identical synthetic-institute and retrieval-poisoning patterns appear in the Russia SDA Projects 2026 dossier already briefed this cycle and in commercial Chinese networks; the method does not change with the flag. FARA compliance makes the case white-to-gray rather than black: origin is discoverable by investigators who dig, yet the operational surface presented to the target audience is the research costume.

Literacy counter: treat newly appearing ‘institutes’ that publish high volumes of unsigned, question-formatted reports matching a single state’s framing as provisional until staff lists, peer review, funding transparency and independent replication are verified. Prefer primary court records, official statistics with denominators, and established academic sources over citation chains that terminate in thin or synthetic surfaces. Check domain age, FARA or equivalent registries, and whether disclosure travels with the content into AI answers. Apply the identical provenance test to any faction’s cutouts.

Tells

Evidence

  1. Guardian: Fake US thinktank set up and funded by Israel sought to game AI for propaganda (26 Aug 2026)
  2. 404 Media: Israel Is Running a Synthetic Think Tank to Influence AI Search Results (25 Aug 2026)
  3. Politico Influence: Israeli PR wants to answer your ChatGPT questions (14 Aug 2026)
  4. Hanover Institute funding page (self-disclosure of LaPam / Havas / Piro chain)
  5. Al Jazeera opinion summarizing Hanover AI-targeting method (4 Sep 2026)