"From the Russia-Ukraine Conflict to the US-Israel-Iran War: Why is the Militarization of AI Becoming a New Global Concern?"

[Global Times reporter Chen Zishuai, Global Times special correspondent Li Jing] Editor's note: The use of artificial intelligence (AI) tools in the U.S.-Israel-Iran conflict is drawing increasing attention and concern. Reports have emerged that the U.S. and Israel used AI systems to analyze intelligence, create strike lists, and simulate combat scenarios during this military operation. However, there are reports suggesting that issues with the AI systems may be related to the tragedy of an airstrike on a primary school in southern Iran, which resulted in the deaths of over 170 people. Additionally, during this conflict, Western tech companies' data centers in the Middle East were targeted by Iran for the first time. The Russia-Ukraine conflict has been referred to as the "first AI war." From the ongoing Russia-Ukraine conflict to the temporarily halted Gaza war, and now to the current U.S.-Israel-Iran conflict, new characteristics of the militarization of AI are continuously emerging. At the same time, the deep involvement of AI on the battlefield has raised a series of issues, including "automated bias," lack of oversight, and the resulting increase in civilian casualties. All of these raise questions about the ethical boundaries of warfare.

From Assistive to Autonomous

On March 11, U.S. Central Command Commander Cooper released a video on social media stating that AI has become an important tool for the U.S. in selecting targets in Iran. According to the Washington Post, insiders revealed that during the U.S. and Israel's "Epic Fury" operation against Iran, both countries used the "Maven Intelligence System" battlefield intelligence platform developed by the U.S. company Palantir. This system incorporates the AI model "Claude" from U.S. company Anthropic, analyzing confidential data from satellites, surveillance systems, and other intelligence channels to provide real-time target locking and prioritization for U.S.-Israeli military operations in Iran. Reports also indicate that this system was previously used in operations against Venezuelan President Maduro.

The U.S. government has significantly expanded the use of the "Maven Intelligence System" across many military departments. As of last May, over 20,000 U.S. military personnel had used the system.

An individual familiar with Anthropic's collaboration with the U.S. Department of Defense told NBC News that "Claude" is primarily used to assist analysts in classifying intelligence and does not directly provide strike target suggestions. However, within the first 24 hours of the "Epic Fury" operation, the U.S. military struck over 1,000 targets in Iran with the assistance of AI. Facing a combat pace that required processing 42 targets per hour, experts like Sylvia from the Royal United Services Institute indicated that this suggests the determination of strike targets is largely automated.

Previously, in the Gaza Strip, AI also became one of the core tools of Israeli military operations. According to a report by Time magazine, the Israeli military used multiple AI systems to generate target suggestions in its retaliatory actions against members of the Palestinian Islamic Resistance Movement (Hamas). Among them, a system named "Gospel" is responsible for identifying and suggesting buildings and facilities where Hamas personnel might be active. The "Lavender" system is programmed to identify suspected members of Hamas and other armed groups for targeted elimination. Additionally, a system called "Where's Daddy" monitors the movements of targets by tracking their mobile phone signals to lock in strike windows. Regarding the role of AI tools in the Gaza conflict, the Israeli military previously stated that "Gospel" and "Lavender" merely "assist intelligence analysts in reviewing and analyzing existing information," and they do not constitute the sole basis for determining attackable targets, nor do they autonomously select targets for attack.

The Russia-Ukraine conflict has been described by many media outlets and scholars as the "first AI war." Some experts predict that by the end of 2024, AI tools will mainly play a supportive and informational role on the Russia-Ukraine battlefield; whether AI-driven artillery systems or AI-guided drones, most are still in the testing phase and have not achieved full autonomy. However, a report by Time magazine in March of this year indicates that this situation is rapidly changing, with various aspects of the Russia-Ukraine conflict becoming increasingly automated.

Most shocking is the large emergence of autonomous drones. These drones are equipped with software that can navigate payloads hundreds of miles away and lock onto targets. Enhanced AI quadcopters from Ukraine can autonomously attack Russian soldiers without human involvement in the event of communication failure. Computer vision enables these drones to identify and attack specific targets, even allowing them to fly through windows for precise strikes. Today, Ukraine launches as many as 9,000 drones daily, becoming a primary testing ground for various global weapon manufacturers, who are competing to automate and make autonomous parts of the conventional "kill chain."

Military expert Zhang Junsha, in an interview with Global Times reporters, stated that in the early stages of the Russia-Ukraine conflict, AI technology was mainly used for reconnaissance assistance, target identification, and intelligence analysis, with limited collaboration between AI systems. In the current U.S. and Israeli military strikes against Iran, however, there has been a deep integration of AI with joint command and control systems, achieving real-time data sharing and collaborative operations across six operational domains: land, sea, air, space, cyber, and electromagnetic. In simple terms, from the Russia-Ukraine conflict to the Gaza war, and now to the current U.S.-Israel-Iran conflict, the role of AI technology and tools on the battlefield is undergoing profound evolution: from assistive to core, from localized to all-domain, from passive to autonomous.

AI Infrastructure Becomes a Target for Military Strikes for the First Time

In the recent U.S.-Israel-Iran conflict, a noteworthy new trend has emerged—AI infrastructure has become a target for strikes for the first time. Amazon Web Services announced that three of its data centers located in the UAE and Bahrain were attacked by Iranian drones. Iranian state television reported that the Iranian Islamic Revolutionary Guard Corps launched the attack to "determine the role these centers play in supporting enemy military and intelligence activities." According to Al Jazeera, the Iranian Islamic Revolutionary Guard Corps recently released a new list of strike targets, which includes data centers and offices of several U.S. tech companies in the Middle East, such as Google, Microsoft, and Palantir.

Iran's vigilance towards Western tech companies is not without reason. For a long time, Silicon Valley tech companies have maintained deep cooperative relationships with the Pentagon. On February 28, the U.S. company OpenAI announced that it had reached an agreement with the U.S. Department of Defense to deploy its AI models on the Pentagon's classified networks. Last year, the Pentagon also signed agreements with U.S. tech companies like Google and xAI, planning to fully utilize these companies' technologies and talents to accelerate the adoption of advanced AI technologies by the Department of Defense.

Western tech companies represented by Google and Amazon provide daily services to ordinary people in many countries while also maintaining deep cooperation with the U.S. military. This has raised concerns about whether the civilian AI tools and data of U.S. tech companies will be used in military contexts. Anthropic has clashed with the White House over concerns that the U.S. government might use its AI tools for mass surveillance of the public and the development of autonomous weapons.

Regarding these issues, Zhang Gaosheng, a scholar at the Institute of World Peace and Security Studies of the China Institute of International Studies, told Global Times reporters that there are extremely high risks. He explained that the vast amounts of data from Western commercial cloud storage, including commercial satellite images, communication data, financial transaction records, and social media trajectories, can easily be accessed by the U.S. military during wartime through permissions or technical means, transforming them into the basis for strike decisions. He warned that digital sovereignty has become a new strategic battleground, and data centers, computing power nodes, and communication cables are directly related to national security and the initiative in warfare. Therefore, "we need to accelerate the construction of autonomous and controllable cloud infrastructure, promote localized data storage, and establish backup mechanisms to ensure survival capabilities in extreme situations."

Ensuring Human Control and Accountability is Crucial

The U.S. and NATO are accelerating the militarization of AI. U.S. Secretary of Defense Austin recently urged the country to accelerate the adoption of AI technology to establish an "AI-prioritized combat force." According to Probasco, a senior researcher at Georgetown University's Center for Security and Emerging Technologies, during a series of exercises codenamed "Crimson Dragon," the U.S. Army's 18th Airborne Corps used Palantir's software, achieving target selection efficiency that matched the records previously set by the unit in Iraq. Probasco pointed out that thanks to AI, this corps achieved this effect with only 20 personnel, whereas completing the same task during the Iraq War required over 2,000 personnel.

NATO, which signed a contract with Palantir last year, recently depicted its version of the "Maven Intelligence System" in a video as capable of providing commanders with battlefield oversight akin to video games.

While AI can provide a significant advantage to a country's military, it also raises widespread concerns. From software failing to recognize or misidentifying personnel to fatal accidents caused by autonomous vehicles, there are numerous examples of AI tools malfunctioning. Regarding the dangers of AI militarization, experts have mentioned "automated bias," which assumes that the information provided by AI is accurate and reliable. U.S. AI expert Asaro pointed out that whether humans truly review each target suggested by AI before authorizing strikes is a question. Israeli journalist Abraham has spoken with six intelligence officials involved in the Gaza operations to report on Israel's use of AI tools. He quoted a target-setting official as saying that although they know the "Lavender" system gives incorrect target suggestions about 10% of the time, they still rely on it. An Israeli intelligence officer responsible for authorizing airstrikes recalled that he spent only about 20 seconds confirming whether a target should be struck, and his specific action was to verify whether the target was male. Many scholars and media believe that the high civilian casualties in the Gaza war are closely related to Israel's use of AI tools.

The use of AI technology by the U.S. and Israel in strikes against Iran not only raises numerous ethical and legal issues but also raises concerns that humanity may be losing control over the war machine. Former Chief of Staff of the Israel Defense Forces Kohavi expressed a similar viewpoint in a 2023 interview, stating, "What is concerning is not that robots will control us, but that AI will control our thoughts and replace us without us realizing it."

Time magazine reported that treaties governing armed conflict do not specify the tools used to achieve military effects. Whether the weapons used are crossbows, tanks, or AI-based databases, they are all subject to relevant international law. However, some institutions and scholars, including the International Committee of the Red Cross, believe that the militarization of AI requires a new legal framework for regulation and point out that as AI weapon systems become increasingly advanced, ensuring human control and accountability over these systems is crucial.

Chinese experts have also issued warnings about the risks of AI militarization. Zhang Gaosheng told Global Times reporters that the speed of AI militarization is accelerating, with countries increasingly caught in a race, but a regulatory framework or consensus regarding AI militarization has yet to be formed, such as adherence to humanitarian principles. Meanwhile, due to the "black box" effect of AI militarization, it is impossible to monitor whether conflicting parties adhere to relevant principles, which poses a significant challenge for arms control, and the international community needs to engage in relevant arms control negotiations.