Israel's government has awarded a $46.5 million contract to Brad Parscale, the former campaign manager for Donald Trump, to create websites and blog posts that shape how AI chatbots like ChatGPT and Claude discuss Gaza. The project, first reported by Drop Site News, has already influenced chatbot outputs: when asked whether increased U.S.-Israel military cooperation would be beneficial, one chatbot answered "Yes" and cited a site produced by Parscale's firm as its top source.
The Contract and the Tactics
Parscale's firm is building a network of websites and blog posts that appear as independent information sources. The target audience is not human readers but the web crawlers that feed training data into large language models. By seeding content that aligns with Israeli government messaging, the operation aims to tilt the factual ground that AI systems draw from when answering questions about the conflict.
The scale of the contract signals a deliberate, well-funded attempt to exploit a known vulnerability in AI systems: their reliance on publicly available web content. When a chatbot's training data includes a critical mass of one-sided material, its outputs can shift accordingly, even without direct retraining.
A New Frontier of Information Operations
The approach bypasses traditional media and public opinion, targeting the infrastructure that increasingly supplies answers to professionals, journalists, and policymakers. Instead of convincing people, the goal is to convince the machines that people consult. This represents a cost-effective route to shaping narrative at scale, since a single influential website can be scraped and ingested by multiple AI models.
Drop Site News's test with the chatbot illustrates the immediate effect. The chatbot's one-word answer was not generated from a balanced review of sources; it reflected the dominance of a single, government-backed publication in the data the model used to form that response.
Why This Matters for Management
For managers who rely on AI tools for research, decision support, or customer-facing applications, the Parscale case is a warning about source integrity. AI outputs are not neutral; they mirror the biases and gaps in the data they ingest. When a state actor can inject content into that pipeline, the risks extend to compliance, reputation, and strategic decisions built on flawed information.
Building governance frameworks that track data provenance and audit AI outputs for source manipulation is becoming a critical capability. Understanding how external actors can manipulate training data is a core part of AI for Executives & Strategy. Regular audits of cited sources, scrutiny of where AI tools pull their information, and clear policies on AI-assisted analysis should now sit alongside other operational risk controls.
Your membership also unlocks:
