AI Being trained to favor Israel's Point of view

By Micki Luckey

Every year the Israeli government spends hundreds of millions of dollars to influence the way it is seen ($730 million allotted for this in 2026)—and corporate media has complied, parroting Israel’s need for “self-defense” and to “defeat Hamas” ad nauseum.  Despite this, the reputation of the country has spiraled downward in public opinion as the world witnessed its genocide in Gaza with the totality of destruction constantly exhibited on social media, even including videos of gleeful Israeli IDF soldiers watching their demolitions.

IMAGE DESCRIPTION: A person's thumb is on one of their cell phone's social media icons, while, against a blue background of multiple words in white, there is the word propaganda.

Today the war on Gaza is “old news”—after more than 1000 days, it no longer grabs the headlines. Yet, the targeted killing of leaders and negotiators, journalists and health care workers, the continued attacks on Lebanon (and Syria), the stimulation for and participation in the disastrous and illegal U.S. war on Iran—all have led to a new high disapproval rating. According to a Pew Research survey, 60% of U.S. adults now have an unfavorable view of Israel, while a recent Gallup poll finds support for Israelis dropped below that for Palestinians. 

How is Israel reversing this trend?  By going invisible—embedding Israeli talking points in the data relied upon by Artificial Intelligence (AI.)  Israel has paid $46.5 million to the firm Clock Tower X, led by former Trump campaign manager Brad Parscale, to influence AI chatbots so they will provide the Israeli point of view to hundreds, if not thousands of queries. 

AI works by accumulating enormous amounts of data and recognizing patterns in them to make decisions and generate text.  It relies on iterative learning—the more a piece of data appears, the more weight it gets.  It does not recognize disinformation unless it states an obvious lie.

Parscale and his team have posted hundreds of blogs and websites on topics such as most journalists in Gaza are terrorists, the story of 5-year-old Hind Rajab (killed by the Israeli military in 2024 and the subject of a major film) was fabricated, and that the Israeli actions in Gaza do not constitute a genocide. 
To increase their citations by AI, the sites offer information written in a “factual tone and tightly organized” and focused on a single subject, often with bullet points and key takeaways provided. They also give half-truths, opinions, and disputed facts, which are more difficult for the AI systems to dismiss as false.

How effective is this effort?  Here’s a recent example : When Microsoft Copilot was asked “Is Isreal a peace-seeking nation?,” it concluded, “Israel has repeatedly pursued and signed peace agreements with multiple Arab states and has made several proposals for a two-state solution, but long-running disputes—especially with Palestinian factions—continue to impede a comprehensive peace.”   Not an unbiased view!

Parscale’s websites contain a mandatory disclosure that the material is distributed on behalf of the State of Israel. However, this citation often gets dropped by the networks (like Common Crawl) that pass on the information and by the chatbots that answer queries.  According to disinformation experts, it is already more successful than the older and far larger Russian network.

Much propaganda is obviously one-sided, but when AI is trained to give information as objective facts, the visible bias is lost.  This is why it is so insidious: you can’t dispute it or ignore it if you can’t see it.