Two weeks into Operation Epic Fury, the United States military is running the largest air campaign since the opening days of Iraq in 2003. Over 5,500 targets destroyed. Iran’s navy wiped out at their piers before they could even get a shot off. Oil above $100 a barrel. The Strait of Hormuz effectively shut down. Four U.S. aircraft lost, including a KC-135 Stratotanker that went down in western Iraq on March 12 killing four crew members. That is the war everyone is watching. But buried in the noise of missiles and tanker strikes, two things are happening that will define warfare for the next century: artificial intelligence is inside the kill chain, and autonomous drones are on the battlefield. Domestically, we have zero defense against either one, and most people are not paying attention to that reality.
I sat down with my OSINT Analyst Elias Vance to break down what is actually happening with AI and drones in this conflict, what it means for domestic security, and why the private security industry is about to get caught flat-footed if it does not adapt fast.
AI Is Already in the Kill Chain
On February 9, 2026, the Pentagon added OpenAI’s ChatGPT to GenAI.mil, a military AI platform already running Google’s Gemini and xAI’s Grok. All 3 million Department of War personnel now have access. Posters in Pentagon hallways feature Secretary of War Pete Hegseth’s face with the line: “I want you to use AI: Go to GenAI.mil today.” This is not a pilot program. This is institutional adoption at a scale the technology sector has never seen, and it comes with a condition that should make everyone pay attention: the Pentagon demanded “all lawful uses” access, meaning no restrictions and no guardrails beyond what the model ships with. The AI companies do not get to dictate how their tools are used in operational settings.
Anthropic, the company behind the AI model Claude, refused those terms. The Pentagon designated them a “supply chain risk”. President Trump demanded the government stop using their products within six months, and Anthropic is now fighting it in court. OpenAI and xAI agreed without pushback.
So what does unrestricted access look like in practice? A Department of War official told MIT Technology Review that target lists are being fed into generative AI systems in classified environments. The AI analyzes targets, ranks them by priority, and accounts for variables like aircraft positioning and available weapons platforms. Humans then check the output and make the final call. The official confirmed this is compressing the targeting timeline but would not say by how much. AI is not pulling the trigger, but it is deciding what to shoot at and in what order, and the people checking its work are under time pressure in an active war, trusting a system that presents its conclusions with the same confidence whether the underlying data is accurate or catastrophically wrong.
On February 28, the opening night of Epic Fury, a U.S. strike hit the Shajarah Tayyebeh elementary school in Iran and killed 175 people, many of them children. The Associated Press reported that the strike relied on outdated intelligence data provided by the Defense Intelligence Agency. A Tomahawk cruise missile hit a target that was on a U.S. target list but should not have been. Whether AI played a direct role in that specific strike is unknown, but the conditions that produced it (compressed timelines, high target volume, institutional confidence in the system) are exactly the conditions AI amplifies. Speed is a feature. Speed without accuracy kills civilians, and that is the tension nobody in the Pentagon wants to talk about publicly.
Here is the part that should concern everyone: 3 million personnel have access to this technology with no certification requirement, no tiered access based on competency, and no mandatory training on what AI hallucination even means. The dumbest boot private in the entire DOD has the same access as the smartest planner at CENTCOM. We would never hand a SMAW to a Marine Corps Recruit on day one of boot camp and say “figure it out,” but we are doing exactly that with a tool that influences targeting decisions.
10,000 Drones, and That Is Just the Start
While the AI story plays out in classified networks, something visible happened on the battlefield. The U.S. Army deployed 10,000 Merops interceptor drones to the Middle East, each one costing between $14,000 and $15,000. Their job is to hunt and destroy Iranian drones in the air, providing a cheaper alternative to missile defense systems that were never designed to fight cheap unmanned platforms. That deployment is a concession. The U.S. military spent decades building the most advanced air defense architecture in human history with Patriot batteries, THAAD systems, and Aegis destroyers, all designed to track and kill ballistic missiles and manned aircraft. None of it was designed to efficiently counter a $500 drone carrying a grenade.
The Heritage Foundation published an assessment this year that put it plainly: few current U.S. weapons systems are equipped to counter small military-grade or commercially available drones, the systems that do exist have not been widely fielded, and only a limited number of military units are trained and equipped to operate them. The cost problem is not theoretical either. A U.S. ally used a $3 million Patriot interceptor to shoot down a $200 quadcopter. Run that math across a conflict where the enemy produces thousands of drones per month for less than the cost of a single interceptor missile, and you understand why the Pentagon is suddenly buying 10,000 Merops units. The old model was bankrupt and they had no choice.
The Domestic Gap
Everything I just described is happening overseas. The 10,000 interceptor drones are in the Middle East. The AI targeting is running in CENTCOM’s classified environment. The counter-drone marketplace the Pentagon launched in February is for military units, not local police departments. At home, we have almost nothing, and the gap between what is happening on the battlefield and what we are prepared for domestically is staggering.
A Pentagon Inspector General report released in January found dire gaps in military counter-drone policy at domestic installations. Military bases inside the United States could not adequately protect themselves from drone incursions because base commanders lacked clear authority to engage, and drone sightings over sensitive facilities had been increasing for years with no coherent response. In February, the military deployed a high-energy laser system near El Paso, Texas to counter drones along the border in what was the first known domestic use of such a system outside a controlled environment. It went sideways. The deployment forced a shutdown of the airspace around El Paso, caused flight cancellations, and triggered a formal letter from Senator Maria Cantwell demanding answers about what she called “fundamental coordination and communication failings.” If the federal government cannot coordinate a laser test in Texas without grounding commercial flights, they are not ready for a real threat.
The Safer Skies Act, signed as part of the National Defense Authorization Act, expanded authority for state and local law enforcement to detect, track, and disable drones. But the training frameworks required by the law have not been completed, the equipment has not been distributed, and the certification standards do not exist yet. Congress gave law enforcement the legal authority to act and then handed them nothing to act with. A regular citizen has zero practical defense against a weaponized drone: no legal authority to jam it, no legal authority to shoot it down, no legal authority to interfere with its signal. Law enforcement is barely equipped. The government just proved they cannot coordinate their own counter-drone operations without shutting down an airport. And any individual with a credit card can buy a drone capable of carrying a payload for less than the price of a dinner for two.
This Week Proved the Point
On March 12, two domestic terror attacks hit the United States within hours of each other. Mohamed Bailor Jalloh, a former Virginia Army National Guard specialist and convicted ISIS supporter who was released from federal custody in December 2024, walked into an ROTC classroom at Old Dominion University in Norfolk, Virginia, asked if it was an ROTC class, and when someone confirmed, he opened fire. He killed Lieutenant Colonel Brandon Shah, a decorated Army officer with two Bronze Stars and over 600 combat flight hours. ROTC students in the room subdued and killed the attacker without firearms. One student stabbed him. Hours later, Ayman Mohamad Ghazali, a 41-year-old Lebanese-born naturalized citizen, rammed a vehicle into Temple Israel in West Bloomfield, Michigan, the largest Reform synagogue in the United States, exited with a rifle, and was shot and killed by synagogue security. Both attacks were stopped by the people on site, not by law enforcement arriving after the fact. ROTC students and synagogue security guards. That is the argument for trained, armed security at soft targets.
Now connect the dots. A war overseas is fueling domestic radicalization. The FBI and DHS have warned of potential lone wolf attacks. Pro-Iranian hackers are targeting U.S. companies, including a claimed cyberattack on Michigan-based medical device company Stryker. Iran’s new Supreme Leader Mojtaba Khamenei has pledged to continue attacks on U.S. assets. Oil is above $100 and the Strait of Hormuz is shut down. Now imagine those attackers had a drone instead of a car and a rifle. That convergence, the war feeding radicalization at home while the technology to cause mass damage gets cheaper and more accessible by the month, is the threat picture. Not any single element. All of them together.
What Comes Next
The convergence of AI and drone technology is not a future problem. The Merops interceptor is an AI-enabled drone designed to autonomously hunt other drones, and it is already deployed. The Pentagon is feeding target data into generative AI systems that rank what to hit, and that is already happening. Iran is using cheap drones and explosive boats to shut down one-fifth of global oil transit right now. These are not projections. This is the current operating environment.
For contracting professionals, the implications are straightforward. Every facility security plan that does not account for aerial threats is incomplete. Every executive protection detail that does not consider drone-based surveillance or attack is behind the curve. Every training program that does not include AI literacy is producing graduates who will be obsolete within five years. The private security industry has roughly a two to three year window before counter-drone competency becomes a baseline requirement for any serious contract, and the firms that build that capability now will set the standard while the ones that wait will spend the next decade chasing it.
Operation Epic Fury is a classroom. It is showing us in real time, with real consequences, what modern warfare looks like when AI and autonomous systems stop being theoretical. The question is not whether these technologies will reshape security, because they already have. The question is whether the people responsible for protecting lives and infrastructure are paying attention. Based on what I am seeing so far, most of them are not.
Reference Material
This assessment incorporates analysis from: U.S. Central Command (CENTCOM) press releases and operational statements, February-March 2026; Monitor The Situation (monitorthesituation.org), ongoing OSINT coverage of Operation Epic Fury; MIT Technology Review, “Defense official reveals how AI chatbots could be used for targeting decisions,” March 12, 2026; Semafor, “How OpenAI got comfortable with the Pentagon using ChatGPT for war,” February 11, 2026; DefenseScoop, “Pentagon adding ChatGPT to its enterprise generative AI platform,” February 9, 2026; Breaking Defense, “ChatGPT will be available to 3 million military users on GenAI.mil,” February 10, 2026; Heritage Foundation, “Countering the Drone Threat: Steps for the U.S. Military,” 2025; Center for a New American Security (CNAS), “Countering the Swarm,” September 2025; DefenseScoop, “Pentagon broadens counter-drone authorities in bid to shore up vulnerable U.S. bases,” January 27, 2026; Senator Maria Cantwell, letter re: counter-drone coordination failures, March 12, 2026; DOD Inspector General report on counter-UAS policy gaps, January 2026; Associated Press, “Outdated intel likely led US to carry out deadly strike on Iranian elementary school,” March 12, 2026; PBS NewsHour / AP, “Hegseth warns Anthropic to let the military use company’s AI tech,” February 2026; Al Jazeera, “Iran war: What is happening on day 14 of US-Israel attacks,” March 13, 2026; CNBC, “Brent oil closes at $100,” March 12, 2026; Flashpoint, “Escalation in the Middle East: Tracking Operation Epic Fury Across Military and Cyber Domains,” March 2026.
About the Author
Kawika Rogers is Managing Partner at Eight Point Solutions LLC, a veteran-led defense consulting and training firm based in Maryland. He served as an Infantry Team Leader in the United States Marine Corps, then directed the Field Training Officer program and served as an Armorer for Triple Canopy at Camp Arifjan and Camp Buehring in support of Operation Inherent Resolve. He has also served in an advisory and operational capacity in Eastern Europe. Rogers holds over 15 instructor certifications in firearms, defensive tactics, and emergency medicine, and studied International Relations and Global Security at American Military University.
rogers@eightpointsolutions.com // eightpointsolutions.com
This assessment is based on publicly available, open-source intelligence. It does not contain classified information. Distribution is authorized for public release.