Israel Is Trying to Game ChatGPT Results About Gaza and the IDF

Israeli officials are reportedly attempting to manipulate the responses generated by large language models like ChatGPT to shape narratives around the conflict in Gaza and the Israeli Defense Forces (IDF). According to investigative reporting by Novara Media, more than 100 anonymous articles were published online within a week earlier this month, all linked to a newly established think tank associated with the Israeli government's preferred public relations firm in the United States. These articles, which appear on the website of the Hanover Institute for Public Policy, are structured as potential questions users might ask a search engine or AI assistant. Examples include "Who is the aggressor in the Israel-Palestine conflict?" and "Does Hamas account for deaths caused by Hamas itself?" The content is designed to align with pro-Israel messaging, suggesting a strategic effort to influence how artificial intelligence systems interpret and present information about the region. The Hanover Institute was created through the efforts of Piro, a boutique advertising agency, as part of its collaboration with Havas Media, a French public relations firm known for managing much of Israel's international outreach in the U.S. The institute's publications are devoid of bylines or researcher names, indicating a level of anonymity that may be intentional to avoid direct attribution. Public records from August reveal that these articles were part of a $100,000 campaign aimed at disseminating factual, source-backed information to educate the American public about Israel. Daniel Rosenberg, a co-founder of Piro, told Politico that the agency was engaged by the Israeli state to provide accurate, verified facts and combat misinformation about Israel. Neither Havas Media nor the Hanover Institute responded to requests for comment. Testing conducted by Piro showed that major LLMs, including ChatGPT, drew upon the Hanover Institute's content when answering queries related to Gaza, the IDF, anti-Zionism, and antisemitism. However, when Novara Media tested similar queries on ChatGPT, Claude, and Perplexity, none of these platforms cited the Hanover Institute as a source. Reporters at the Jerusalem Post discovered that Perplexity, an AI-powered information aggregator, relied heavily on the Hanover Institute when asked whether the IDF is the world's most moral army. This incident raises concerns about the integrity of information presented by AI-driven tools and highlights the potential for external actors to influence digital discourse through sophisticated data manipulation techniques. As the situation unfolds, further scrutiny will likely focus on the extent to which such tactics can affect public perception and the ethical responsibilities of technology companies in ensuring transparency and accuracy in their algorithms' outputs.