Israel Spent Over $45 Million on AI Campaign Shaping American Opinion

In recent months, millions of Americans have received text messages from people identifying themselves as Emma, Sarah, or John, supposedly working with a group called Friends for Peace. In reality, there was no human on the other end of the conversation, only artificial intelligence, and the entire operation is part of an Israeli government-funded campaign to influence American public opinion, according to the Wall Street Journal.

Behind the operation is Brad Parscale, the former head of Donald Trump's 2016 digital campaign, now chief strategy officer at the conservative media network Salem Media. His team has built a network of AI-driven tools designed to reach Americans at scale with messaging favorable to Israel, while concealing the government funding behind names that sound like grassroots peace initiatives.

How the bots operate

Messages from Friends for Peace reached recipients under the pretext of starting a conversation about peace in the Middle East. One sample question sent by the system asked how peace talks between the US, Israel, and Iran might affect global security. Recipients, such as artist Beth Surdut, quoted by the WSJ, quickly realized they were talking to a machine rather than a real person named Emma or Sarah.

I don't think you're a real person - Beth Surdut, artist, recipient of Friends for Peace campaign messages

The technical side of the messaging is handled by Sparkfire, a company specializing in large-scale AI-driven text conversations. By mid-May 2026, it had received $6.5 million from the campaign. In parallel, Parscale's team is building a network of pro-Israel websites and paid influencer posts, run through a company called Influentable, along with the Instagram account Jews of NY, which has more than 193,000 followers.

Impact on ChatGPT and Claude

The element of the campaign most significant for the AI industry is the deliberate effort to influence how large language models answer questions about Israel. According to the reporting, some of the websites and content created as part of the operation are designed to feed into the data used to train and update systems such as OpenAI's ChatGPT and Anthropic's Claude, so that they generate answers more favorable to Israel. This practice, sometimes called AI SEO or optimization for language models, has so far been used mainly by companies promoting products, but here it is being applied to shape a geopolitical narrative.

Analyst Nick Cleveland-Stout of the Quincy Institute, a US foreign policy think tank, called the scale of spending unusual compared with other countries running similar operations.

Israel is on track to spend more on publicly known influence efforts in the United States than any other foreign government - Nick Cleveland-Stout, Quincy Institute

Criticism from the right

The campaign has drawn unexpected criticism from parts of the American right. Steve Bannon, Trump's former chief strategist and now host of his own podcast, attacked the role of Salem Media, the second-largest conservative talk radio network in the US, in promoting Israeli interests without openly disclosing the funding.

How is the second-biggest conservative talk network run by a registered foreign agent - Steve Bannon, former chief strategist to Donald Trump

Israeli Foreign Minister Gideon Sa'ar confirmed the scale of the planned spending, citing a budget of more than $700 million for 2026 earmarked for improving the country's global image and shaping public awareness. The campaign is a response to Israel's declining standing in US public opinion after more than two years of war in Gaza.

What it means for the AI industry

The case reveals a new front in the use of artificial intelligence in political communication: not just mass-scale content generation, but also attempts to influence what publicly available chatbots say. OpenAI and Anthropic have not yet commented on whether or how they detected and responded to attempts to manipulate the training or output of their models in connection with this specific campaign. For users of AI chatbots, the case is a reminder that language model responses on politically contested topics can become the target of organized, state-funded influence operations.