5 Uncomfortable Truths About AI in Warfare

In our daily lives, artificial intelligence has become a tool for the mundane—an LLM to help polish an email or a predictive algorithm to curate a playlist. But as world leaders and experts gathered in Geneva for the June 2026 Informal Exchange on AI in the Military Domain, a much darker narrative emerged. The 2026 AWS Diplomacy Report describes a world of “accelerated warfare” where the window for international regulation is rapidly narrowing, leaving humanity at the mercy of the “automation of violence.”

Between June 15 and 17, 2026, a tension-filled dialogue unfolded between those rushing toward a future of drone swarms and those warning of a “black box” of machine learning that even its creators cannot fully explain. The consensus was chilling: technology is moving at a breakneck pace, and we are losing our grip on the kill chain.

Here are five uncomfortable truths about the current state of AI in warfare.

1. The Tech Giants Have Abandoned Their Red Lines

For years, Silicon Valley marketed itself as the ethical counterweight to the traditional defence industry. That facade has shattered. The tech and arms sectors are fusing into a singular military-industrial complex where the distinction between civilian and military technology is fading away.

Google has removed its pledge not to design AI for weapons. Meta and OpenAI have reversed previous bans on military use. Most damning is the role of Anthropic, a company that has built its entire brand on “safety” and “alignment.”

“Anthropic—despite its safety branding—deployed Claude across a US nuclear-weapons laboratory and into Palantir’s military environment.” — AWS Diplomacy Report, Finding #2

This shift has created a staggering “due diligence gap.” Corporate human rights policies have been strategically rewritten to govern suppliers rather than customers or military end-users. By deferring entirely to the judgment of home governments, these companies have effectively outsourced their conscience to the state.

2. The Myth of Surgical Precision

The defence industry markets AI as a tool for “surgical precision” and enhanced civilian protection. However, civil society groups like the International Committee for Robot Arms Control (ICRAC) and Privacy International argue that the primary logic is actually “speed and output” over quality.

The goal is to “squash the kill chain”—compressing the time it takes to identify and strike. In practice, feeding more data into these systems creates an “exponential increase” in potential targets rather than better accuracy. We are already seeing the results of “algorithmic objectivity” being used as a veneer for mass violence. AI-driven target generation has reportedly outpaced the capacity for genuine human review in operations in Gaza, Iran, and Lebanon, as well as reported US operations in Venezuela.

“If we build systems that can only escalate, accelerate, and intensify armed conflict, then we are setting our world down a path to perpetual war.” — ICRAC

3. The Danger Begins Long Before the Battlefield

We often focus on the moment a drone fires, but the Office of the United Nations High Commissioner for Human Rights (OHCHR) argues that we must look at the entire “Life Cycle” of harm. Human rights risks are “hardcoded” during the pre-design and data collection phases.

This “pipeline of harm” includes:

• Data Bias: Systems are trained on data that reproduces the “noise” and prejudices of its context—sexism, racism, and ableism—turning data-intensive systems into tools that reinforce colonialism and patriarchy.
• Infrastructure Targets: Because military operations now rely on commercial clouds and satellite-internet constellations, civilian digital infrastructure has become a valid military target, placing communities on the front lines of modern conflict.
• Digital Dehumanisation: By reducing humans to data points within a “black box” dataset, the technology erodes the moral engagement required to take a life.

“The human rights risks of AI do not begin at deployment; they are often shaped much earlier through decisions about the need for the system, the purpose of the system, the data used, [and] the design choices made.” — OHCHR

4. A Global Majority Says “No” to Machine-Led Killing

A chasm has opened between the dictates of the state and the public conscience. While a small clique of military powers—the U.S., Russia, Israel, Burundi, and the DPRK—voted against UN Resolution 80/58, the global population is increasingly terrified.

Recent survey data of over 6,000 adults across eight nations reveals a “surprising” mandate for human control:

• 79% of Americans want humans to make lethal decisions.
• A majority in countries including Brazil, China, and South Africa support a total ban on AI weapons that target people.

Under the “Martens Clause,” international law is supposed to be guided by the “dictates of the public conscience.” Currently, major powers are ignoring this clear global mandate, arguing that regulation would “stifle innovation” while the public begs for safety.

5. When AI Touches the “Red Button”

The most terrifying prospect discussed in Geneva is the integration of AI into Nuclear Command, Control, and Communications (NC3). Experts warned that “algorithmic manipulation” or “hallucinations” could trigger a nuclear escalation before a human commander can even process the data.

Because machine learning systems operate as “black boxes,” decision-making processes are not fully understandable to their operators. This creates a high risk of “automation bias,” where humans simply rubber-stamp a machine’s recommendation. To prevent this, many are calling for a “non-delegation pledge,” yet the rush for “force multiplication” makes the integration of AI into nuclear decision-making feel like an impending, algorithmic accident.

Conclusion: Choosing Peace Over “Fate”

The 2026 AWS Diplomacy Report makes it clear: technology is not an inevitable force of nature. It is a choice—a reflection of our values and the kind of future we are willing to build. As we move toward a world of “rubber-stamping killing,” we must decide if we will allow the speed of technology to outpace our humanity.

The Stop Killer Robots Interfaith Working Group poses the question that should haunt every tech executive and diplomat: Are we building a future of perpetual war, or will we have the political will to keep the “kill chain” human?

Are you monitoring the intersection of automated state warfare, tech stacks, and international policy?

• See my stance on AI development and governance here: aviperera.com/stance
• Explore more deep dives into machine-comprehensible infrastructure and AI safety: aviperera.com/aisafety

Avi Perera

Avi is a researcher educated at the University of Cambridge, specialising in the intersection of AI Ethics and International Law. Recognised by the United Nations for his work on autonomous systems, he translates technical complexity into actionable global policy. His research provides a strategic bridge between machine learning architecture and international governance.