收听本期播客
阅读正文
The increasing integration of Artificial Intelligence (AI) into modern warfare has sparked significant debate, raising profound ethical questions about accountability and the impact on civilians. Many experts now contend that companies developing military AI systems should be reclassified as defence contractors, rather than simply technology firms, to address these growing concerns.
The discussion often references the ‘fog procedure’, an informal military tactic where soldiers fire into darkness, believing in an unseen threat. This historical practice, which essentially permits violence through a deliberate lack of clear vision, finds a disturbing modern parallel in AI warfare. Critics argue that the ‘darkness inside the algorithm’ – the opaque nature of AI decision-making processes – is itself a deliberate design choice. This opacity, they suggest, allows for deniability when harm occurs and makes the resulting violence appear an inevitable consequence of automated systems.
Recent conflicts have highlighted the urgent need for clarity. The conflict in Gaza, for instance, has been controversially labelled the first major ‘AI war’, with AI systems playing a central role in identifying potential targets. These systems rapidly process vast quantities of data to assign individuals a probability of being combatants. A tragic illustration of this danger occurred in Minab, Iran, where a US strike resulted in 168 deaths, predominantly children, at an elementary school. The building had been a military base a decade earlier, but this crucial update was missing from the targeting system. While the weapons used were precise, the intelligence guiding them was dangerously flawed and outdated.
The sheer speed at which AI can generate thousands of potential targets means that human operators often have only seconds to ‘verify’ these suggestions. This rapid pace effectively reduces human oversight to merely approving machine decisions, a process akin to ‘rubber-stamping’. This situation severely complicates the traditional chain of command, blurring accountability across engineers, military commanders, and corporate suppliers. International humanitarian law (IHL) demands careful processes and verifiable reasoning for military operations, requirements that the opaque nature of many AI systems often fails to meet.
Major technology giants, including Palantir, Google, Amazon, Microsoft, and OpenAI, are deeply embedded in military operations, providing systems vital for targeting. However, these private companies largely operate beyond the reach of international laws designed for states. Current regulations, such as the EU AI Act, frequently exempt military applications, creating a substantial gap in oversight. Critics assert that these firms are central to the modern military-industrial complex and should therefore be subjected to the same stringent regulations as traditional defence contractors, including export controls and robust liability frameworks.
This complex landscape makes it exceedingly difficult to assign full responsibility when AI-driven strikes lead to civilian casualties. It fundamentally alters the decision-making process in warfare, carrying profound implications for human rights and the future of international law.