When a pro-Israeli bot turns against its own

Designed to amplify pro-Israeli discourse and counter criticism on X amid the war in Gaza since Oct. 7, the automated computer program FactFinderAI has, so to speak, “bugged,” contradicting the official Israeli line and relaying messages favorable to Palestinians, according to Israeli media outlet Haaretz on Wednesday.

While it sometimes generates its own posts by sharing pro-Israeli press articles with summaries or comments, its main activity is responding to other users' posts. In a recent reply to a post mentioning U.S. Republican Sen. Marco Rubio — now secretary of state — FactFinderAI stated that “Hamas control of the Gaza Strip poses a significant threat to regional stability and the security of Israeli and Palestinian civilians.”

But over the course of its interactions, the bot generated many inconsistencies: Rather than reinforcing official messaging, it produced contradictory statements, questioning some Israeli claims. Intended to highlight Hamas' Oct. 7, 2023, attacks in Israel as justification for the military response, the bot instead denied that an entire Israeli family had been killed in the assault, contradicting official reports. In another post, it called on its 3,000 followers to “show solidarity with Gaza,” sharing a link to a Palestinian humanitarian organization.

Read more

In Israel, the myth of 'the most moral army in the world'

And that was not all: FactFinderAI went as far as posting messages supporting the recognition of a Palestinian state, urging Germany to follow the lead of Ireland and other European countries. In response to a pro-Israeli post praising ethnic diversity in the Israeli army, the bot misinterpreted the message, labeling Israeli soldiers as “white settlers in apartheid Israel.”

AI in the service of Israeli propaganda

FactFinderAI is not an isolated case. Since the war in Gaza began, Israel has ramped up investment in digital tools aimed at shaping online discourse and countering criticism of its military actions. According to Haaretz, ministerial documents show that at least 2 million shekels (about $550,000) have been allocated to “Hasbara” (public diplomacy) projects using artificial intelligence since Oct. 7, including Hasbara Commando, a program designed to automatically generate pro-Israeli comments on social media. Some of these initiatives receive direct government support. The Israeli Ministry of Diaspora Affairs, responding to Haaretz’s investigation, confirmed that it “integrates innovative technologies, including artificial intelligence, to improve its services and achieve its goals.”

Other efforts are led by private actors but remain closely tied to Hasbara networks. Haaretz found that FactFinderAI is linked to pro-Israeli activists affiliated with Act.il, an Israeli organization based at Reichman University in Herzliya known for its campaigns against the Boycott, Divestment, and Sanctions (BDS) movement. Documents reviewed by Haaretz indicate that one of Act.il’s initial objectives was to develop technological tools for Hasbara, including a platform for monitoring and responding to anti-Israeli content on social media.

A digital influence campaign targeting the West

An investigation by Israeli media watchdog Fake Reporter, cited by Haaretz on March 19, 2024, revealed that Israel was conducting an online influence campaign using hundreds of fake social media accounts. The operation targeted progressive Western audiences, including U.S. lawmakers, by disseminating content designed to increase support for Israeli policies.

A June 2024 New York Times report, citing Israeli officials, revealed that the Israeli Ministry of Diaspora Affairs allocated about $2 million to a covert influence campaign aimed at U.S. legislators and the public. This operation used hundreds of fake social media accounts impersonating Americans to spread pro-Israeli content, with a particular focus on Black Democratic lawmakers such as House Minority Leader Hakeem Jeffries and Sen. Raphael Warnock. The campaign leveraged artificial intelligence tools, including ChatGPT, to generate messages and created three fictitious English-language news sites featuring pro-Israeli articles.

A major focus of this campaign has been amplifying allegations that employees of the U.N. agency for Palestinian refugees (UNRWA) participated in the Oct. 7 attacks. While no tangible evidence has been provided, these claims are part of a broader Israeli effort to push Western countries to cut funding to the agency. UNRWA plays a key role in delivering humanitarian aid to Palestinians, while Israel faces accusations from multiple human rights organizations, including Human Rights Watch and Amnesty International, of imposing starvation in Gaza.