Militarised AI

Last week, a trove of leaked documents offered a glimpse into the role that large technology companies have played in Israel’s war on Gaza. Israel’s military campaigns across Palestine and neighbouring countries have long offered the raw material that tech firms needed to build their experiments in surveillance and algorithmic warfare to scale: kill-chain data. Yet the past fifteen months of war offered Silicon Valley an unparalleled opportunity to refine its products. It happened just in time for a new era of militarised AI.

After Hamas militants breached Israel’s border with Gaza on 7 October 2023 and killed more than 1100 civilians and soldiers, the Israeli army mobilised 360,000 reservists. Some of them held senior positions at Google, Microsoft or Amazon in Tel Aviv. It seems that a few had been working on services with obvious military applications, from cloud computing to large language models. Israel’s constant aerial bombardments and the start of a protracted ground invasion were creating a deluge of data. The army needed more computing power and storage. The mobilised employees asked their companies for help.

Google gave the IDF access to Vertex, which applies AI algorithms to data sets, and Gemini, an AI assistant that the IDF wanted to repurpose to sift through text or audio and recommend operations. Microsoft allowed the Israeli army’s use of Chat GPT-4, accessed through Microsoft Azure, to increase twenty-fold. Amazon expanded the cloud storage for data informing military operations across Gaza. Soldiers described accessing classified information for lethal bombardments on Amazon cloud servers as ‘making an order from Amazon’.

The news may be shocking but it isn’t surprising. Yossi Sariel, the head of Israel’s Intelligence Corps Unit 8200 from 2021 to 2024, has written of ‘empowering the relationship between government and big data companies’. Beyond signing million-dollar contracts with big-tech, Sariel aimed to remake intelligence units in the image of AI start-ups through consulting and joint training. Some soldiers in Unit 8200 worked in what their officers referred to as ‘AI factories’, building up speech-to-text software for Palestinian Arabic, tagging phrases to be flagged in keyword searches, or making classified systems more user-friendly. Some surveillance databases were given nicknames like ‘Facebook for Palestinians’.

In a document self-published on Amazon in 2021, Sariel dreamed of fantastic and ethically dubious AI applications, such as a ‘military Waze’ that could automatically pinpoint targets and guide lethal operations. It seems that Sariel did what he could bring his vision to life. Microsoft was paid millions to host private development meetings and professional workshops for units across Israel’s intelligence agencies. According to +972 Magazine, Microsoft Azure employees were referred to as ‘people who are already working with the unit’ as if they were active soldiers. By the start of the Gaza war, the IDF had a host of AI-assisted targeting systems to guide its bombing spree, and was developing even more.

The partnerships were mutually beneficial. In exchange for technology and expertise, the Israeli army offered companies it worked with access to an unlimited supply of data. Many of the deals contravened Silicon Valley’s doveish mission statements and human rights policies. Some companies denied involvement in Israeli military operations. ‘This work is not directed at highly sensitive, classified or military workloads relevant to weapons or intelligence services,’ Anna Kowalczyk, the external communications manager for Google Cloud, told reporters who asked about the company’s dealings with the IDF in April 2024.

With Trump’s return, other companies have simply abandoned those talking points and embraced militarism. Company boards seem undeterred by employee protests or arrest warrants issued by the International Criminal Court against senior Israeli officials. Late last year, OpenAI, Meta and Anthropic signed their first deals with the US Department of Defense. These partnerships could deepen following the news that the Chinese company DeepSeek launched an AI chatbot at a fraction of the cost of other AI assistants. Perhaps in a bid to galvanise sponsors for their own dubious business models, some Silicon Valley technocrats warned that China may win the AI arms race.

It remains to be seen what all this investment in militarised AI will deliver, beyond more violence. The Israeli military’s love affair with Silicon Valley has done little to shore up regional security. As a number of Israeli ex-security officials have pointed out, none of the applications developed and deployed in recent years prevented Hamas’s attacks on 7 October 2023. Sariel resigned in September 2024, citing his ‘personal responsibility’ for the ‘intelligence and operational failure’. In fifteen months of war, military experiments with Chat GPT-4 and other generative AI applications from Microsoft and Google failed to achieve any of Israel’s stated war aims in Gaza.