The Israeli government is pushing a strategy involving a budget of hundreds of billions of won to influence AI chatbots to respond to Israel-related information in a specific manner.
Imagine this: tonight, you ask an AI chatbot, “Summarize the Israel-related issues in the news right now.” The chatbot explains it step-by-step in a very plausible tone. You trust that the AI is providing the most ‘neutral’ response based on vast amounts of internet material. But what if that answer is the result of a ‘script for learning’ meticulously crafted with a specific government budget?
Recently, controversy has erupted over reports that the Israeli government is deploying a new form of digital PR strategy using artificial intelligence. It goes beyond simply showing advertisements to people; it seeks to change the very ‘lens of data’ through which AI views the world.
Why It Matters
When we have questions, we now turn to AI chatbots like ChatGPT before using search engines. AI responses are already becoming a guide for our thoughts. If someone secretly draws a map for this ‘guide’ to head only in a certain direction, we risk unconsciously thinking only in that direction.
The strategy being pursued by the Israeli government goes beyond the level of creating national PR videos. Because AI learns from internet data, this is an attempt to fill AI’s ‘knowledge repository’ itself with information favorable to them. This poses a major homework assignment for us to seriously contemplate regarding democracy and the reliability of information in the digital age.
The Explainer
Let’s compare the principle of AI responses to a ‘giant library.’ The librarian (AI) reads countless books (internet data) and answers the questions of patrons (users). Until now, it seemed like the librarian was fairly referencing all the books in the library.
However, the PR agency ‘Clock Tower X LLC,’ hired by the Israeli government, is trying to secretly flood this library with books containing specific content. Source: Responsible Statecraft These books are not even standard promotional materials; they are elaborately disguised as if written by authoritative ‘research institutes’ or ‘think tanks.’ Source: Haaretz
Simply put, they are trying to manipulate ‘where AI gets its information.’ If AI models like ChatGPT, Gemini, and Claude reference these ‘fake research institute’ materials during their training process, the probability of them providing answers reflecting Israel’s stance naturally increases. This method is professionally known as ‘GPT framing.’ Source: Ynetnews To use an analogy, it’s like making a librarian constantly read only books from a certain perspective, eventually leading them to answer patrons according to that perspective.
Where We Stand
It is reported that the Israeli government is investing a massive budget of over half a billion shekels (hundreds of billions of won) for this project. Source: Ynetnews Clock Tower, which signed a $6 million contract, is focusing particularly on creating content that can reach Gen Z, who use TikTok, Instagram, and YouTube. Source: DefendDemocracy.press
One notable point is that AI is actually providing different answers depending on our location. Research results show that when asked about Israel from within Israel, the chatbot uses language emphasizing security, the right to self-defense, and the war on terror, while adopting a more cautious stance on moral questions. Source: Ynetnews
Of course, rumors that OpenAI and the Israeli government signed a deal to directly modify AI’s responses have been revealed to be false. Source: Yahoo News However, this strategy is evaluated as far more subtle and threatening because it manipulates the ‘data environment’ that AI feeds on and grows in, rather than manipulating the AI directly.
What’s Next
Attempts by governments or giant corporations to use AI to represent their national or corporate interests will continue to increase. It will become much more difficult than it is now to discern the truth of information.
Many experts, including reporter Omer Benjakob, question how effective these attempts will actually be and warn that such methods could contaminate the information ecosystem. Source: Haaretz What can we do to become wise users in the digital age? The most important thing is to have a critical attitude and ask yourself once in a while, “Where was this information learned?” when listening to AI responses. This is because AI is both a mirror for the questions we throw at it and a reflection of data that someone may have intentionally placed.
References
-
[ChatGPT asked famous Israelis questions - here are their answers The Jerusalem Post](https://www.jpost.com/j-spot/article-738709) - Making Israel’s case to ChatGPT and Grok: Hasbara meets AI in multi-million dollar PR push - YouTube
-
[Israel wants to train ChatGPT to be more pro-Israel Responsible Statecraft](https://responsiblestatecraft.org/israel-chatgpt/) -
[Report: Israel to spend over half a billion shekels turning ChatGPT into public diplomacy tool Ynetnews](https://www.ynetnews.com/tech-and-digital/article/rj00kxqzaxx) -
[Ask ChatGPT about Israel — your location may change the answer Ynetnews](https://www.ynetnews.com/tech-and-digital/article/skilyoscwx) - Making Israel’s case to ChatGPT and Grok - Haaretz
-
[Israel didn’t sign a deal with ChatGPT. Here’s what actually … Yahoo News](https://uk.news.yahoo.com/israel-didnt-sign-deal-chatgpt-233300413.html) -
[Israel wants to train ChatGPT to be more pro-Israel DefendDemocracy.press](https://www.defenddemocracy.press/israel-wants-to-train-chatgpt-to-be-more-pro-israel/)
- They signed a direct contract with OpenAI to modify the system
- They intentionally create data for AI to reference through fake research institute websites, etc.
- They block users from using AI
- The elderly
- Scholars worldwide
- Gen Z
- It answers identically everywhere in the world
- The content or tone of the answer may vary depending on the user's location
- AI never answers political questions