Who Gets to Tell the Military “No”?
As the Pentagon increasingly integrates artificial intelligence (AI) into military operations, a critical question emerges: Who gets to tell the military “no”? The stakes have risen significantly as recent negotiations highlight a concerning shift in how AI companies interact with military directives. Anthropic, a notable AI firm, chose to uphold ethical standards regarding AI use in warfare but lost a major government contract as a result. In contrast, OpenAI embraced the Pentagon’s more lenient approach, favoring compliance over caution. Investigative reporting from The Intercept sheds light on the implications of this decision.
Understanding OpenAI’s Commitment
The agreement between OpenAI and the Pentagon explicitly stipulated the types of AI models the Department of Defense (DOD) would utilize: “OpenAI models that are designed for national security use cases and have minimal refusal rates.” This troubling phrase indicates that the military is looking for systems that seldom, if ever, oppose its orders. A Justice Department lawyer, representing the Pentagon, initially confirmed this language but later retreated from that position, reflecting the tension surrounding the ethics of AI in military applications.
The Hierarchical Nature of Military Orders
At first glance, the military hierarchy justifies a system where orders are followed without question. Officers, with their extensive training and battlefield experience, generally make decisions that are in line with military objectives. However, this raises the question: If soldiers are obligated to follow orders, should machines be exempt from this directive?
The Duty to Disobey
In a strictly legal context, military personnel have the obligation to refuse illegal orders. According to U.S. military law, members are required to comply with lawful directives while refusing unlawful ones. This principle is enshrined in the DOD’s Law of War Manual, emphasizing adherence to international law and ethical standards.
This “duty to disobey” isn’t merely theoretical. Following WWII and the Nuremberg Trials, where many Nazi soldiers were prosecuted for committing war crimes under the guise of following orders, it became clear that individuals hold moral and legal responsibility regardless of their position within a command structure.
Are AI Systems Capable of Moral Judgments?
With the introduction of AI systems into military operations, the question arises: Can the rationale behind the duty to disobey be extrapolated to machines? Unlike human soldiers, AI systems lack the ethical framework needed to discern the nuances of legal versus illegal orders. As highlighted by Rebecca Crootof, a technology law expert, while AI can be programmed to follow established laws, the interpretation of law is often context-specific—something AI struggles to navigate.
The Ambiguity of Military Orders
Interpreting legality on the battlefield is rarely a black-and-white issue. Determinations of legality hinge on subtle factors, such as the behavior of enemy combatants or the situational context. While AI may have access to legal documents and battlefield data, its effectiveness in making nuanced judgments is questionable.
Striking a Balance
Given the complexities involved, the potential solution lies not in relying solely on AI for decision-making but rather in creating a hybrid model. By programming AI systems to flag ambiguous scenarios—such as the presence of non-combatants or uncertain combatant behavior—for human review, military operations maintain ethical oversight while leveraging technological advantages.
As these AI technologies become more embedded within the military framework, public transparency regarding their operation and limitations is crucial. The open acknowledgment of these systems’ capabilities—and their potential ramifications—will help foster accountability.
For more insights on this topic, please refer to the complete article Here.
Swati Sharma
Vox Editor-in-Chief
Image Credit: www.vox.com
