US military AI targeting investigation faces intense scrutiny after a missile strike on an elementary school in Minab killed 175 people, most of them schoolgirls. CENTCOM says advanced AI speeds analysis while humans still approve strikes. Early findings point to outdated data, weak oversight, and a deadly gap between databases and the real world. The head of US Central Command, Admiral Brad Cooper, confirmed that American forces now use a range of AI tools to scan data and speed decisions in the war with Iran. He said the systems help leaders cut through noise and act faster than enemies. He also said a human will always make the final call on what to hit. But a preliminary probe, reported by The New York Times, ties a 28 February Tomahawk strike in Minab to a “targeting fiasco.” Investigators believe the strike used old coordinates from the Defense Intelligence Agency. The school, bright blue and pink with clear sports fields, split from a nearby base in 2016. It stayed flagged as an active target in military databases.

US military AI targeting investigation: What we know so far

The Minab strike and legacy data

The Minab blast killed 175 people, including 150 schoolgirls and staff, according to early counts. The site looked civilian and had served children for years, but records did not keep up. That gap between a database and the ground turned deadly. The US military AI targeting investigation now centers on three weak points:

Stale intelligence that mislabels civilian sites

Poor database hygiene that fails to remove old targets

Rushed human review inside a fast “kill chain”

Did AI misidentify the school?

Experts say we do not yet know. Dr. Craig Jones of Newcastle University told The Times that AI may have failed to identify the school as a school and could have tagged it as military. That is one possible link in the chain. Another is human. A person still had to accept the target and launch the strike. This is the core tension. AI can compress work that once took days into seconds. But if the inputs are wrong or the checks are thin, the machine makes a faster path to a bad decision. Speed without fresh facts and careful review can magnify harm.

Speed versus safety in the modern kill chain

Admiral Cooper argues speed gives an edge. He says AI helps leaders decide faster than an enemy can react. Critics argue that faster is not safer when the data is dirty and the review is brief. The US military AI targeting investigation now looks at whether speed eroded basic safeguards.

Wider fallout and pressure

The Iranian Red Crescent says nearly 20,000 civilian buildings and 77 health sites have been damaged in the fighting. China’s Defense Ministry warned that letting algorithms steer life-and-death choices risks a “technological runaway.” The Trump administration pushed back. After a legal fight with Anthropic over ethical limits, a Pentagon spokeswoman said the military will not be “held hostage by Silicon Valley ideology.” Against that backdrop, officials say the US military AI targeting investigation will examine how algorithms flagged targets, how old data stayed live, and how commanders checked civilian status before firing.

What should change before the next strike

Five fixes that reduce civilian harm

Audit and clean target databases now. Remove or re-verify any site last checked years ago, starting with schools, hospitals, and clinics.

Demand fresh, multi-source confirmation. Match AI alerts with new satellite images, human reports, and signals before approval.

Build a strong “no-strike” shield. Force AI and humans to check against up-to-date protected site lists, with hard blocks that require senior overrides.

Slow down at the last step. Set a minimum “think time” and peer review for strikes near populated areas, even when AI shows high confidence.

Track and test the models. Log model inputs, outputs, and confidence. Red-team them against known edge cases like schools next to bases.

Publish civilian harm reviews. Share methods and lessons learned with independent monitors to restore trust and improve practice.

Human control that means something

“Humans in the loop” only works if people have time, training, and authority to say no. Real control means clear rules, visible risk flags, and freedom from pressure to match machine speed. It also means accountability when the process fails, whether the error began with a person, a database, or a model.

The data problem is the AI problem

AI is only as good as the data and labels it sees. If a school and a base share a fence, the model can get confused unless it learns from precise, current examples. If the target list is wrong, the best model will still speed the wrong answer. Clean data, updated maps, and constant re-validation are not optional add-ons. They are the foundation.

Precision needs context

The Minab strike was “picture-perfect” in aim, according to reports, but still hit the wrong building. That shows a hard truth: precision without context can be deadly. Modern warfare needs both. AI can help find context faster, but only if teams build systems that spot civilians first and force caution, not only speed. In the end, technology will not carry the moral weight. People will. As the US military AI targeting investigation moves ahead, it should deliver more than blame. It must set strict guardrails, clean the data, and put meaningful human control back at the center before the next decision crosses the line between warfighting and tragedy. (pSourse: https://www.trtworld.com/article/ce171f6a1aa5)

For more news: Click Here

FAQ