Gaming the Machine

There is a question that gets asked less often than it should: if Big Tech's platforms have been suppressing Palestinian voices, what has been filling the space those voices left behind? The answer is not silence. It is manufactured content, coordinated influence operations, and a systematic campaign to train the AI systems that hundreds of millions of people now use as their primary source of information — to produce the answers that governments want them to produce.

This chapter is about how that works. It is about who is paying for it, who is building it, and what it means for the future of truth itself.

The $6 Million Operation: Gaming ChatGPT Through the Back Door

In September 2025, newly filed documents with the U.S. Department of Justice under the Foreign Agents Registration Act confirmed what many had suspected but could not yet prove. The Israeli government — operating through its Foreign Ministry and a $145 million public diplomacy initiative called Project 545 — had hired a network of American firms to conduct one of the most ambitious AI influence operations ever documented.

At the center of the operation stands Brad Parscale — the digital architect of Donald Trump's 2016 presidential campaign, the man who hired the now-notorious data firm Cambridge Analytica, and now the chief strategy officer of Salem Media Group, a conservative Christian broadcasting network whose stakeholders include Donald Trump Jr. and Lara Trump. Parscale's firm, Clock Tower X LLC, registered as an agent of the Israeli government on September 18, 2025, under a $6 million contract from German media giant Havas Media Network, which was itself operating on behalf of Israel's Ministry of Foreign Affairs.

The goal was explicit and the method was new. Clock Tower would produce a minimum of 100 "root creative assets" per month — videos, graphics, audio, written content — with at least 80 percent targeted at Gen Z audiences across TikTok, Instagram, YouTube, and podcasts. The target: 50 million paid impressions per month. But what made this operation different from conventional PR was one specific clause buried in the contract's statement of work. Clock Tower would deploy, in the filing's own language, "websites and content to deliver GPT framing results on GPT conversations." (Source: Responsible Statecraft, "Israel Wants to Train ChatGPT to Be More Pro-Israel," October 2025, https://responsiblestatecraft.org/israel-chatgpt/; Sludge, "Israel's New U.S. Influence Campaigns Target TikTok, Churches, and ChatGPT," October 2025, https:// readsludge.com/2025/10/06/israels-new-u-s-influence-campaigns-target-tiktok-churches- and-chatgpt/)

Let me explain what that means in plain language. AI systems like ChatGPT are trained on vast amounts of text from across the internet. They learn what is true, what is contested, and what framing to apply to complex topics by processing billions of web pages, articles, and social media posts. If you flood the internet with a particular version of events — through hundreds of

websites, millions of social media posts, and coordinated content designed to rank highly in search engines — you are not just shaping public opinion directly. You are shaping what AI learns. You are, in effect, poisoning the well from which the next generation's information will be drawn. Clock Tower would also use a predictive AI platform called MarketBrew to optimize content for algorithmic visibility on Google and Bing — ensuring that the pro-Israel narrative would be what both human users and AI training systems encountered first, and encountered most.

This is not a direct payment to OpenAI. Let me be precise on that point, because the rumor circulating online — that Israel paid OpenAI directly — is inaccurate, and this book deals in documented facts. Snopes investigated and confirmed: there is no evidence of a contract between Israel and OpenAI. What is documented, in legally filed government disclosures, is something more sophisticated and more difficult to challenge: a government-funded operation to game the training environment of AI itself. (Source: Snopes, "Israel Didn't Sign a Deal With ChatGPT. Here's What Actually Happened," January 2026, https://www.snopes.com/news/2025/12/22/israel-contract-chatgpt/; Business and Human Rights Resource Centre, https://www.business-humanrights.org/en/latest-news/israelopt-clock-tower-x-allegedly-hired-by- israeli-government-to-create-pro-israel-content-training-chatgpt-and-targeting-facebook-instagra m-tiktok-youtube-incl-cos-response-and-non-responses/)

When the Business and Human Rights Resource Centre contacted OpenAI, Meta, Google, Alphabet, and Havas Media requesting comment on these allegations, only TikTok responded. OpenAI said nothing. That silence is its own answer.

OpenAI and the Military Contract That Wasn't — Until It Was

There is a second story about OpenAI and Israel that has nothing to do with influence operations. It is about weapons. And it is documented by the Associated Press.

In February 2025, the AP published an investigation based on internal company documents, data, and exclusive interviews with current and former Israeli and company officials. The findings were stark. Following Hamas's October 7, 2023 operation, Israel's use of OpenAI technology — purchased through Microsoft's Azure cloud platform — spiked to nearly 200 times its pre-genocide level. The amount of data the Israeli military stored on Microsoft servers doubled in the first months of the genocide alone, reaching more than 13.6 petabytes by July 2024 — roughly 350 times the digital memory needed to store every book in the Library of Congress. OpenAI's AI models were being used directly in the intelligence and targeting pipeline of an active military campaign. (Source: Associated Press, "As Israel Uses US-Made AI Models in War, Concerns Arise About Tech's Role in Who Lives and Who Dies," February 2025; Fortune, https://fortune.com/ 2025/02/19/israel-microsoft-openai-raises-questions-powerful-tech/)

"This is the first confirmation we have gotten that commercial AI models are directly being used in warfare," said Heidy Khlaaf, chief AI scientist at the AI Now Institute and a former senior safety engineer at OpenAI herself.

OpenAI's response to this revelation was to claim it had no direct partnership with the Israeli military. That claim deserves scrutiny. In early 2024 — quietly, without announcement — OpenAI rewrote its terms of service to remove language that explicitly prohibited military use of its technology. The ban on military applications was simply deleted. What had been a line in the

sand became an open door. And through that door walked the Israeli Military Forces, purchasing OpenAI models through Microsoft's Azure platform, using them to process surveillance intelligence and support targeting decisions. The company changed the rules, and then pointed to the rules as proof of innocence. That is not a defense. That is a business strategy.

Google's Gemini: The Whistleblower's Evidence

Google has consistently maintained that its AI products are not used for weapons or surveillance in violation of its ethics policies. In February 2026, a former Google employee filed a confidential whistleblower complaint with the U.S. Securities and Exchange Commission telling a different story.

According to the complaint — reviewed by The Washington Post — in 2024, a customer support request arrived at Google's cloud computing division from an address linked to an Israeli Military Forces account. The customer was attempting to use Google's Gemini AI to analyze aerial drone surveillance footage and had encountered a bug where the software was occasionally failing to identify drones, soldiers, and other objects in the footage. Google's cloud unit customer support staff responded with technical assistance and conducted internal tests to help resolve the issue. A second Google staffer — who allegedly worked on the Israeli Military’s Google Cloud account — was copied on the support ticket. (Source: Washington Post, "Whistleblower Says Israeli Military Contractor Used Google's Gemini AI," February 2026, https://www.washingtonpost.com/technology/2026/02/01/google-ai-israel-military/)

The whistleblower alleged that this assistance violated Google's own AI principles, which at the time stated the company would not deploy AI for weapons applications or surveillance violating internationally accepted norms. Google disputed the characterization, claiming the support ticket was too small in commercial value to constitute "meaningful" AI usage. That is a remarkable defense. The question of whether a drone surveillance tool violates ethics does not change based on the size of the monthly billing statement.

Separately, documents from November 2024 show the Israeli military actively requesting access to Google's Gemini AI platform to develop its own AI assistant capable of processing military documents and audio recordings — expanding its capabilities months into an active genocide. Records show Google employees noticed a surge in the Israeli Military’s access requests for AI tools after October 7, with internal documents revealing that one employee warned that delaying expanded access might push Israel toward Amazon, Google's cloud competitor. (Source: The Defense Post, "Google Supporting Israel With AI Tools for Military Use," February 2025, https://thedefensepost.com/2025/02/04/google-israel-ai-military/; Common Dreams, https:// http://www.commondreams.org/news/big-tech-gaza-genocide)

Google's response — like OpenAI's, like Microsoft's — follows the same pattern we have documented throughout this book. Deny. Minimize. Update the policy to close the gap between what was done and what was permitted. Repeat.

The Pattern Behind the Pattern

What emerges from these three stories — the Clock Tower influence operation, the OpenAI military usage spike, and the Gemini whistleblower complaint — is not three separate scandals. It is one coherent strategy, operating simultaneously on two fronts.

On the military front, AI companies provide the tools that make targeting faster, more automated, and more lethal. They do this through direct contracts, through cloud platforms, and through the quiet removal of the ethical guardrails that once made such use impermissible.

On the information front, governments use AI's own training process as a weapon — flooding the internet with manufactured content designed to teach AI systems what to believe, what to suppress, and whose suffering to treat as legitimate. The result is an information environment in which the same AI tools that help identify bombing targets are simultaneously being trained to describe those bombings in language favorable to the people ordering them.

This is the full architecture of AI-enabled control. It does not only kill. It also shapes what you believe about the killing. And both functions are being provided, knowingly or not, by the same companies — OpenAI, Google, Microsoft — whose quarterly earnings reports you will read next week in the business pages of your newspaper of record, between the advertisements.

OpenAI was asked to comment on the influence operation designed to game its training data. It did not respond. That silence tells you everything you need to know about whose interests these companies have decided to serve.

---

--- If Chapter Ten is about how AI is trained to manufacture consensus for adults, Chapter Eleven is about what happens to the children who grow up inside platforms designed not to inform but to addict. The algorithm that rewards extremist content for adults is the same algorithm that recommends self-harm content to teenagers. The platform that censors Palestinian voices is the same platform that the New Mexico jury found had knowingly harmed children's mental health. These are not separate stories. They are the same business model applied to two different vulnerable populations.