Israel developing ChatGPT-like tool for surveillance of Palestinians

Israel’s military is developing an advanced artificial intelligence tool similar to ChatGPT, trained on Arabic conversations collected through surveillance of Palestinians living under occupation. 

The tool is being developed by the Israeli army’s cyber warfare Unit 8200, which is programming the AI to understand colloquial Arabic by feeding it vast amounts of phone calls and text messages between Palestinians, gathered through surveillance.

Three Israeli security sources confirmed the existence of the AI tool to the investigating outlets. The model was still in training last year, and it is unclear whether it has been deployed yet or for what purpose. However, sources stated that the tool’s ability to process large volumes of surveillance data and “answer questions” about specific individuals would provide a major advantage to the Israeli military.

Several sources indicated that Unit 8200 has used smaller-scale machine learning models in recent years. One source mentioned, “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.”

An Israel Defense Forces (IDF) spokesperson declined to comment on the new AI tool but emphasized that the military uses various intelligence methods to “identify and thwart terrorist activity by hostile organizations in the Middle East.”

Unit 8200 previously used AI tools like The Gospel and Lavender during the war on Hamas, which played a significant role in identifying potential targets for strikes. The unit has been using AI for nearly a decade to analyze intercepted communications, categorize data, recognize patterns, and make predictions.

After the launch of ChatGPT in November 2022, the Israeli military set up a team to explore how generative AI could be adapted for military use, according to former intelligence officer Chaked Roger Joseph Sayedoff. However, OpenAI, ChatGPT’s parent company, rejected the unit’s request for direct access to its large language model (LLM) and refused to integrate it into the military’s system.

Sayedoff also highlighted another challenge: existing language models could only process standard Arabic, not the spoken dialects of Arabic. This prompted Unit 8200 to develop its own model. One source noted, “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.”

Unit 8200 began recruiting experts from private tech companies in October 2023, with Ori Goshen, co-CEO of AI21 Labs, confirming that his employees participated in the project as part of their reserve duty.

The challenge was to compile all the spoken Arabic text the unit had gathered and centralize it. Eventually, the training data for the model included about 100 billion words, including conversations in Palestinian and Lebanese dialects.