Israel's $45M AI Influence Campaign Is Targeting US Voters — and Your Chatbot

Israel has spent over $45 million on an AI-powered US influence campaign run by Trump's 2016 digital chief Brad Parscale, using fake personas to send AI-generated texts and seeding web content to manipulate how ChatGPT, Claude, and other AI assistants answer questions about Israel, according to a Wall Street Journal investigation. The operation, operating under a 'Friends for Peace' banner, sits inside a $700 million Israeli 2026 'consciousness shaping' budget and represents the first documented nation-state AI training-data influence operation at production scale.

Israel's $45M AI Influence Campaign Is Targeting US Voters — and Your Chatbot The Wall Street Journal revealed Israel has spent over $45 million on an AI-powered US influence campaign run by Trump's 2016 digital chief Brad Parscale. The operation uses fake personas 'Emma' and 'Sarah' to send AI-generated texts under a 'Friends for Peace' banner. The more novel tactic: seeding web content to reshape how ChatGPT, Claude, and other AI assistants answer questions. It sits inside a $700M Israeli 2026 'consciousness shaping' budget — the first documented nation-state AI training-data influence operation at production scale. The Wall Street Journal published a detailed investigation this weekend into Israel's $45 million AI-powered influence operation aimed at American public opinion — and the most unsettling part isn't the millions of text messages. It's the systematic effort to reprogram what ChatGPT /compare/chatgpt-vs-claude and Claude /glossary/claude say about Israel. The mechanics: Brad Parscale, Trump's 2016 digital campaign chief now serving as chief strategy officer at Salem Media, has been paid roughly $1.5 million per month under a contract now exceeding $45 million. The campaign operates under a "Friends for Peace" banner and sends AI-generated text messages to US cellphones from fake senders named "Emma" and "Sarah." It targets younger conservatives in particular. But the text message operation is the visible layer. <<<BOLD The more novel tactic is AI chatbot /glossary/chatbot manipulation.<<<BOLDEND The campaign seeds websites and content specifically engineered to be ingested by the web crawlers that train and update large language models. The goal: shape how ChatGPT, Claude, Gemini /glossary/gemini , and other AI assistants answer questions about Israel, Gaza, settlers, and the broader conflict. It's SEO poisoning, but for AI training /glossary/training data instead of Google rankings — and the asymmetry is staggering. A few million dollars in content generation can alter the answers that hundreds of millions of users receive from their AI assistants. The $45 million sits inside a broader Israeli budget line of $700 million for 2026 "consciousness shaping," according to the WSJ report. Six firms were retained over the past year. Steve Bannon has publicly questioned Salem Media's foreign-agent-linked leadership, creating an unusual fault line on the right between pro-Israel hawks and America-first skeptics. <<<BOLD This is the first documented case of a nation-state running a production-scale AI training-data influence operation.<<<BOLDEND It's not hypothetical anymore. It's not a "what if" in an academic paper. It's active, funded at eight figures, and being run by the same digital strategist who helped put Trump in the White House. The AI industry has spent two years debating alignment, safety, and hallucination /glossary/hallucination . This story suggests those conversations are missing the threat vector that's already live: nation-states aren't waiting to hack the models. They're flooding the training data. Get AI news in your inbox Daily digest of what matters in AI. Key Terms Explained Chatbot /glossary/chatbot An AI system designed to have conversations with humans through text or voice. Claude /glossary/claude Anthropic's family of AI assistants, including Claude Haiku, Sonnet, and Opus. Gemini /glossary/gemini Google's flagship multimodal AI model family, developed by Google DeepMind. Hallucination /glossary/hallucination When an AI model generates confident-sounding but factually incorrect or completely fabricated information.