Investigation uncovers Israeli military’s AI tool for spying on Palestinians
text_fields
DUBAI: The Israeli military is reportedly developing an advanced artificial intelligence tool, similar to ChatGPT, designed to analyze Arabic conversations obtained through surveillance of Palestinians living under occupation.
A joint investigation by The Guardian, Israeli-Palestinian publication +972 Magazine, and Hebrew-language outlet Local Call uncovered details of the project, which is being developed by the Israeli army’s secretive cyber warfare division, Unit 8200. The AI tool is being trained to understand colloquial Arabic by processing vast amounts of phone calls and text messages between Palestinians obtained through surveillance.
Three Israeli security sources familiar with the project confirmed its existence to the investigating outlets. Although the model was still undergoing training last year, it remains unclear whether it has been deployed or how it is being used. However, sources suggested that the AI’s ability to rapidly analyze large volumes of surveillance material and generate answers about specific individuals could provide significant advantages to the Israeli military.
The investigation further revealed that Unit 8200 has previously used smaller-scale machine learning models. One source emphasized the power of AI in intelligence gathering, stating: “AI amplifies power; it’s not just about preventing shooting attacks. I can track human rights activists, monitor Palestinian construction in Area C (of the West Bank). I have more tools to know what every person in the West Bank is doing. When you hold so much data, you can direct it toward any purpose you choose.”
When approached for comment, an Israel Defense Forces spokesperson declined to address questions about the AI tool. However, the military stated that it “deploys various intelligence methods to identify and thwart terrorist activity by hostile organizations in the Middle East.”
Unit 8200 has previously developed AI tools such as The Gospel and Lavender, which were used during the war against Hamas to identify potential targets for strikes and bombardments. The division has also spent nearly a decade using AI to analyze intercepted communications, categorize data, recognize patterns, and make predictions.
Following the public launch of ChatGPT’s large language model in November 2022, the Israeli army established an intelligence team to explore how generative AI could be adapted for military purposes. However, OpenAI, the parent company of ChatGPT, rejected Unit 8200’s request for direct access to its large language model and refused to allow integration into the unit’s systems.
One challenge in developing the AI tool was the need to process spoken Arabic in various dialects. Existing language models primarily focus on standard Arabic, making them insufficient for the military’s needs. “There are no transcripts of calls or WhatsApp conversations on the internet. It doesn’t exist in the quantity needed to train such a model,” said one source.
To overcome this limitation, Unit 8200 began recruiting experts from private tech companies as reservists in October 2023. Ori Goshen, co-CEO and co-founder of Israeli tech firm AI21 Labs, confirmed that his employees participated in the project while serving as reservists.
A key challenge for Unit 8200 was consolidating vast amounts of intercepted spoken Arabic text into a centralized database. A source stated that the AI model’s training data eventually comprised approximately 100 billion words, including conversations in Lebanese and Palestinian dialects.
While AI models offer significant intelligence advantages, Goshen acknowledged their limitations, stating: “These are probabilistic models—you give them a prompt or a question, and they generate something that looks like magic, but often the answer makes no sense.”
Experts have raised concerns about the ethical implications of such AI-driven surveillance. Zach Campbell, a senior surveillance researcher at Human Rights Watch, described AI tools as “guessing machines,” warning that “these guesses can end up being used to incriminate people.”
Campbell and Nadim Nashif, director and founder of the Palestinian digital rights advocacy group 7amleh, highlighted the potential misuse of personal data in AI training. Campbell noted: “We are talking about highly personal information, taken from people who are not suspected of any crime, to train a tool that could later help establish suspicion.”
Nashif argued that Palestinians have become test subjects in Israel’s efforts to develop and weaponize AI for controlling occupied populations. “This is a grave and continuous violation of Palestinian digital rights, which are human rights,” he said, emphasizing that such technologies serve to reinforce occupation and surveillance systems.