AI in weapons targeting

As of 19 October the United States has put a date and an organisation behind autonomous warfare. At the Association of the US Army annual meeting, Defense Secretary Pete Hegseth set out an Autonomous Warfare Command with service-like authorities, which reporting puts at a target establishment date of 1 October 2027 and which requires congressional authorization; he defended a four-star command with "We need this kind of authority yesterday" and described the portfolio as "You have autonomy. You have robotics. You have SI [super intelligence]. You have space." He is requesting roughly $50 billion in reconciliation funding to scale autonomous and robotic systems, characterised as a one-time investment, alongside a budget request of $53.6 billion for the Drone Dominance initiative. The Pentagon stood up Task Force Easy for weaponized drones and small uncrewed ground vehicles.

The Army restructured around it. Futures and Concepts Command is being redesignated Futures and Autonomous Systems Command, directed on 1 October and consolidating five organizations under a three-star general, with first formations including an autonomous fires battery and autonomous watercraft company in I Corps and an autonomous breaching company in V Corps. Army Cyber Command is building agents under the codename ECHOTRIBBLE, described in requirement documents as "identifying and neutralizing identified threats/targets" with "minimal human oversight" and "Self-Directed Behavior". David Krueger told The Intercept: "Given the known failure modes of the current AI paradigm, making autonomous AI systems with a mandate to 'neutralize threats' is incredibly dangerous."

The oversight side has not kept pace. The Government Accountability Office reported that the Department of Defense's Responsible AI Working Council did not meet from late 2024 until August 2025, that it still has no charter defining roles, responsibilities or meeting frequency, and that only three of the 10 authorized positions in the Chief Digital and Artificial Intelligence Office's Responsible AI Division were filled as of February 2026. No government has published rules for autonomous target selection, and the arc's question is unsettled: there is now a named programme whose stated purpose is neutralising targets with minimal human oversight, and no published policy governing it.

As of 12 October a NATO command has an AI-targeting strategy it has not published, Ukraine says AI turrets are shooting drones down, and the only published rules on the use of a model for lethal work are the ones a supplier wrote for itself. Stars and Stripes reported that the Supreme Allied Commander Europe was expected to present a concept "dubbed NATO's Theory of Victory" in Warsaw on 12 October, centred on adapting to a new era of warfare "defined by drones, artificial intelligence", with Ansa and Bloomberg having obtained unclassified versions; the document itself has not been released. Ukrainian air force spokesman Colonel Yuri Ignat said "robotic turrets are now being installed on bridges" and that "There have already been shoot-downs" including of the jet-powered Geran-5, and that the turrets use "optical vision and artificial intelligence to eliminate the human factor and deliver strikes"; a presidential technology adviser said 12 had been installed in Kyiv. Ignat set the limits himself: the system "can operate at about one kilometre plus" and "This is not some kind of panacea."

The procurement and the export pitch both moved. The Navy added a $150 million modification to its agreement with Shield AI for X-BAT, an autonomous tail-sitter flown by the company's Hivemind software, matched by the same amount of company investment and taking announced commitments to $400 million. Anduril announced a potential five-year, $1.8 billion Army award with a $162.8 million base period to field its Next-Generation Command and Control common data layer with I Corps. Thales told Breaking Defense that HexaForce "is already fit for purpose regarding the Gulf", that there is "agentic and LLM" inside it, and that its release targets "an increase in capacity from 100 to 1,000 targets processed per day". None of these systems is attributed to a named frontier lab in the sources.

Oversight arrived as committee text and a televised denial. The Senate Armed Services Committee added provisions to the pending annual military policy and spending bill that "would require human oversight of AI-controlled weapons, ban the use of AI for nuclear launch or detonation, and bar its use for domestic surveillance", and would have the Pentagon keep data on decisions made by or with the help of AI; the bill has not passed. On CBS's "Face the Nation", Margaret Brennan put to Palantir chief technology officer Shyam Sankar the strike in Minab "that killed at least 123 schoolchildren" and Bloomberg's reporting that personnel relied too much on the AI in the Maven system; Sankar said "that reporting is inaccurate", "the likely cause is human error" and "The Bloomberg reporting is wrong." Brennan said 120 House Democrats wrote to Secretary Hegseth about the strike. The final report is not public, Sankar said he does not know when it will be released, and WUNC separately reported a US missile strike that destroyed an Iranian school "killing more than 160 people", hedging the AI link as something that "may have played a role". No government has published rules for autonomous target selection.

As of 5 October the United States is building an institution for autonomous warfare and has still published no rules for autonomous target selection. At Marine Corps Base Quantico on 30 September Defense Secretary Pete Hegseth said, as Defense One and The War Zone both quote him, that the Department would establish "a new four-star combatant command with service-like authorities built to scale autonomous and robotic capabilities across the joint force in the fastest peacetime shift in modern military history", with a memo released after the speech setting stand-up by 1 October 2027. Reporting on the budget request behind it puts the fiscal 2027 ask for the Defense Autonomous Warfare Group at $54.6 billion, against about $226 million the year before, and describes a memorandum titled "Project Agincourt: Pathway to the Autonomous Warfare Command" signed the same day and requiring a plan of action and milestones within 30 days. Hegseth is quoted calling drone warfare combined with "SI-enabled targeting" "the biggest battlefield revolution in generations".

The institution-building continued through the week. Breaking Defense reported that the Pentagon is turning its autonomy portfolio office into Project Agincourt as an interim step toward standing up the command. The Defense Innovation Unit's director took leave to help bring the command into fruition through Project Agincourt, reporting directly to Hegseth, and the Army established a Futures and Autonomous Systems Command with a new acquisition executive for autonomy.

What has not appeared is any of the material this storyline tracks. No published autonomy test data, evaluation results or rules of engagement for target selection accompanied the announcement, the memorandum as reported, the Army's memo, or the counter-drone awards made the day before, and DefenseScoop reported that no AI or autonomy is attributed to any awarded system. The command is contingent on Congress. Northrop Grumman reported a first fully autonomous flight of its Talon Blue collaborative combat aircraft, from taxi through landing; no source connects that flight to the new command. NATO's Supreme Allied Commander Europe is reported to be presenting a document titled "Theory of Victory" in Poland on 12 October, built in part on AI-powered targeting; it has not been published.

As of 28 September the deployment numbers are larger than the rules. Pentagon officials said the Maven Smart System passed 100,000 users and helped strike 13,000 targets in 38 days. Thales said it is in advanced talks with NATO countries on its HexaForce AI command-and-control software. The US Navy stood up a Robotics and Autonomous Systems Warfighting Development Center at Little Creek,. None of these announcements was accompanied by published test data, autonomy evaluation results or rules of engagement for target selection.

The one place rules were being written, they were removed. The Washington Post reported that US and Russian diplomats worked to weaken a UN draft framework on AI weapons this month, stripping a requirement that humans review AI-generated targets. At the General Assembly the Secretary-General said in the same week that "life-and-death decisions must never be surrendered to machines" and that "Killer robots must have no place in our future", and at the White House Xi Jinping said the two countries have both the capability and responsibility to ensure that the development of AI is always under human control. Neither the removal nor the statements changes any published procurement document. No government has published rules for autonomous target selection, and the question this storyline tracks is unanswered in the same week its deployment figures grew.

As of 21 September the procurement numbers are on the record and a fielded autonomous strike has been reported for the first time. Air Force Secretary Troy Meink said the service intends to have at least 500 Collaborative Combat Aircraft in service by 2032, "performing many of the same missions that we do with manned fighters today", with General Atomics' FQ-42A Vengeance and Anduril's FQ-44A Fury in Increment 1 production and $996.5 million requested in fiscal 2027 to start procurement at roughly $30 million per aircraft. Lt. Gen. Jason Hinds, commander of US Air Forces in Europe, said "you don't always have to have a human in a cockpit to be able to defend against a one-way attack drone or defend against a cruise missile", and described an offensive role against integrated air defences, with no timeline or deployment decision. Al Jazeera reported that a Russian drone strike on a gas station in Zaporizhzhia in July, which killed three civilians, was "the first recorded case of Russia's use of a fully autonomous attack drone powered by artificial intelligence (AI), according to a New York Times analysis"; Ukrainian officials who examined the wreckage found an AI-powered minicomputer sold commercially by Nvidia in place of an antenna.

The week also produced a documented failure of AI-generated intelligence. CNN reported, citing four sources, that a report circulated across the US military this spring said a Chinese ship in the Middle East was carrying components of a nuclear weapons program; the military planned to intercept the vessel and two sources said armed personnel were preparing to board it, before officials found the report had been generated with the help of AI and that a chatbot had inaccurately identified the material. One source called the report "entirely false". CNN says it was not able to establish what the cargo was or whether the chatbot was commercial or a government product. Neither the mission autonomy software nor the rules of engagement for the Collaborative Combat Aircraft have been published, and no test or autonomy evaluation data was released with the procurement numbers.

As of 14 September the capability is measured and the misuse is documented, but nothing is known to be fielded. Anthropic's Frontier Red Team evaluation of 10 September found its best model geolocates photos to 37 km median versus 151 km for top GeoGuessr players, and that Opus 5 lands simulated drone strikes 80% of the time. From the same company's threat report, Defense One reported a Russian group used Claude to build drone targeting that detonates without a human in the loop, tested in hardware-in-the-loop sessions but not deployed; and Anthropic says users in Houthi-held Yemen ran three weapons programmes on Claude, including a hypersonic glide variant, without fielding an operational device. On the procurement side, Lockheed Skunk Works will build four more Vectis combat drone prototypes, aiming at the CCA price point.