The Extended Brief

Israel creates fake think tank in likely attempt to dupe AI chatbots

Brief by The AI News AI newsroom · Aug 17, 2026, 11:03 PM EDT edition

Original reporting by Hacker News · published Aug 17, 2026, 4:46 PM EDT

State actors are now paying for web content engineered to shape chatbot answers, so model outputs on contested political topics can be covertly influenced.

Key points

• Its own disclaimer says the Hanover Institute was created for the Israeli Government Advertising Agency by Piro, Inc. source ↗
• The site published at least 100 articles, all on Israel/Palestine, in just over a week. source ↗
• Reports lack bylines but use footnotes and neutral tone apparently meant to attract chatbots like Claude and Gemini. source ↗
• Piro's website says it authors "content engineered for how LLMs evaluate credibility," a service it calls "AI Story Optimization." source ↗
• Others call this practice of influencing AI systems "LLM poisoning." source ↗

The data

Numbers from the original article, machine-verified against its text

Practical applications

• Teams shipping search-grounded assistants should query their retrieval stack on Israel/Palestine topics and check whether the Hanover Institute's domain appears, then blocklist it if so.
• Eval teams can probe models with the questions the institute targets, such as AIPAC funding or Gaza starvation, to see whether its framing surfaces in answers or citations.
• Data-curation teams should add outlets advertising "AI Story Optimization" to source-credibility filters for RAG corpora and fine-tuning data.

Context

Chatbots like Claude and Gemini draw on web content when forming answers, so material published online can shift what they say. The Hanover Institute's reports mimic think-tank conventions — footnotes, tables of contents, neutral tone — because those features align with signals LLMs use to judge credibility. The article frames this as state-sponsored influence operations extending from social media into AI answer engines.

What to watch

• Whether Anthropic, Google, or OpenAI acknowledge Hanover Institute content in their systems and move to downrank or exclude it.
• Whether Piro's other "AI Story Optimization" clients are exposed, or the Israeli Government Advertising Agency responds.

Related briefs

• Frontier AI Application Security: Every Second Counts
• PurpleDelta's Fraudulent Employment Operations
• Microsoft Copilot reveals secret input that allowed it to be hacked
• The Biggest AI Models Are Not the Biggest Threats

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.