AI as Military Advisor: Who Decides in the War of the Future?

What happened?

Artificial intelligence has crossed a critical line: from being a logistical support tool, it has become a military advisor capable of recommending tactical actions in real time. An exclusive eBook from MIT Technology Review, compiled by journalist James O'Donnell and published in June 2026, documents how various armies are integrating AI models into their command centers. The material, which brings together six reports published between April 2025 and April 2026, reveals that systems like the Pentagon's Project Maven —originally launched in 2017 to analyze drone imagery— have evolved into target recommendation platforms. NATO, for its part, has launched the 'Alliance AI Command' program that tests algorithms in war simulation exercises. According to the report, these systems have already been used in real conflicts, although operational details remain classified.

Why is it important?

The speed and complexity of modern conflicts —with autonomous drones, synchronized cyberattacks, and electronic warfare— have overwhelmed human processing capacity. AI can analyze terabytes of data from sensors, intercepted communications, and satellite imagery in seconds, offering courses of action that theoretically reduce the risk of tactical errors. However, delegating critical decisions to machines raises deep ethical dilemmas: who is responsible if a faulty recommendation causes civilian casualties? Can an AI distinguish between a combatant and a civilian in a chaotic environment? The MIT Technology Review report underscores that, although AI is presented as an 'advisor,' in practice its influence can be decisive. A Stanford University study cited in the eBook notes that in simulations, human commanders accepted AI recommendations 89% of the time, even when they contradicted their initial judgment.

Consequences and challenges

• Risk of algorithmic bias: Training data can reflect historical prejudices or intelligence errors. For example, Project Maven was criticized in 2020 for biases against certain ethnic groups in its image analysis. The eBook documents how these biases persist and can lead to discriminatory recommendations, such as prioritizing targets in areas of a specific ethnic majority.
• Lack of transparency: Many models are 'black boxes' whose decisions cannot be fully explained. One cited case is NATO's 'Aegis AI' system, which in a 2025 test recommended an attack on a hospital based on erroneous troop movement patterns. Engineers could not determine why the algorithm made that decision, sparking an internal debate on accountability.
• Arms race: Countries like the US, China, and Russia are competing to develop advanced military AI. According to a report from the Center for a New American Security (CNAS) mentioned in the eBook, China has invested over $100 billion in military AI since 2020, while Russia has deployed recommendation systems in Syria. This competition increases the risk of inadvertent conflicts, such as the 2025 incident where an Iranian drone and a US drone nearly collided due to conflicting recommendations from their respective AIs.
• Human control: Initiatives like the 'Declaration on the Responsible Use of AI in the Military Domain,' signed by 42 countries in 2023, aim to keep a human in the decision loop. However, the eBook reveals that implementation is uneven: while the US requires human review for lethal attacks, other countries like Israel have fully automated certain target selection processes.

What readers should know

AI will not replace human commanders in the short term, but its role as an 'advisor' is already shaping strategies. The MIT Technology Review eBook warns that the speed of recommendations can pressure officers to accept them without sufficient scrutiny, creating a 'tyranny of the algorithm.' Moreover, technological dependence creates vulnerabilities: a cyberattack that manipulates input data could trigger catastrophic decisions. In 2025, a NATO exercise demonstrated that a data poisoning attack caused the AI to recommend attacks against allied positions. The international community has yet to reach a consensus on binding limits, although forums like the Geneva Convention are discussing updates for the digital age. The eBook concludes that transparency and independent auditing are urgent, but national interests hinder any agreement.

“AI is not neutral: it reflects the biases of its creators and the data it was trained on. In the military domain, that can have lethal consequences,” the report states.

For more details, it is recommended to consult the full eBook from MIT Technology Review, available in PDF and ePub.