Israel accused of AI influence campaign

Israel accused of AI influence campaign

Israel allegedly spent tens of millions of dollars on a digital campaign aimed at shaping how artificial intelligence systems respond to questions about the war in Gaza.

The report claims that Brad Parscale, a former campaign manager for US President Donald Trump, and his company, Clock Tower X, secured a $46.5 million contract from the Israeli government to create an online network promoting Israeli viewpoints. The campaign reportedly sought to influence AI platforms such as ChatGPT, Claude, Gemini and Perplexity.

According to documents cited in the report, Clock Tower X created hundreds of articles on websites covering the Gaza war, Israel’s military operations and US-Israel relations. The strategy focused on producing large amounts of online content that search engines could index and AI systems could later use when generating answers.

The report described the approach as a form of “LLM poisoning,” in which coordinated online content attempts to shape the information that large language models encounter during web searches, data collection or training.

Websites reportedly linked to the campaign, including Allyvia.org, FactSignal.org and Paxpoint.org, published material supporting stronger US-Israel military cooperation, defending Israel’s actions in Gaza and disputing allegations against the Israeli military.

Drop Site News also reported that some AI-powered services have already referenced content from the network. In one test, Perplexity reportedly cited Allyvia.org in response to a question about US military cooperation with Israel, while Microsoft’s Copilot also referenced websites associated with Clock Tower X.

The report further claimed that content from the network entered Common Crawl, a widely used archive of web pages that developers and researchers rely on to build and train AI systems. The 10 websites allegedly connected to Clock Tower X appeared in Common Crawl records hundreds of times between January and June, potentially increasing their visibility to AI systems and influencing future AI-generated responses.