The Franchise Model

In December 2025, the US Treasury sanctioned a Colombian mercenary network recruiting fighters for Sudan's RSF paramilitary. The action targeted two private firms — A4SI and Maine Global Corp SAS — and read, at the time, like a straightforward enforcement action against rogue operators.

The United States again calls on external actors to cease providing financial and military support to the belligerents — United States Department of the Treasury

Read against everything that has happened since, that sanction was not an isolated case. It was the enforcement edge of a system the US has been building across three domains simultaneously — a system that is not outsourcing but franchising sovereign offensive capability. The distinction matters. Outsourcing delegates a task: the state hires a contractor to do something the state could do itself. Franchising delegates the authority to conduct sovereign operations under the state's legal cover, doctrinal framework, and target selection, while the franchisee bears the operational execution, the legal liability, and the diplomatic deniability. The state keeps who gets authorized, what gets targeted, and what oversight applies. The firm keeps the risk. This franchise model is not one policy. It is one structural pattern clicking into place across three domains. In offensive cyber, the architecture is now explicit. Trump's March 2026 "Cyber Strategy for America" laid down the doctrine. ^[1]

We will unleash the private sector by creating incentives to identify and disrupt adversary networks and scale our national capabilities. — Donald Trump

A September 2025 White House program had already allowed vetted private firms to conduct "hack back" operations against adversaries — a quiet first iteration. ^[2] The August 2026 National Security Presidential Memorandum formalized and expanded it: vetted US private companies can now conduct offensive cyber operations — disruption, manipulation, or destruction of foreign information systems — against transnational criminal organizations, with DOJ and DHS written approval and supervision, and a $1 million bond requirement. ^[3] The state provides the legal cover. It does not provide immunity. The firms carry the legal exposure if an operation goes wrong. In AI-powered combat targeting, the same pattern holds. Project Maven — the Pentagon's AI targeting system, built and operated by Palantir — generated lethal target lists used in US strikes on Iran, compressing what the military calls the kill chain "from hours to seconds." ^[4] More than a thousand targets were struck in 24 hours. The Pentagon has since signed contracts with eight tech companies — Amazon, Google, Microsoft, Nvidia, OpenAI, Oracle, SpaceX, and Reflection — to integrate generative AI into classified military networks, with the stated goal of building "the United States military as an AI-first fighting force." ^[5] It launched a $100 million prize challenge for private firms to build voice-controlled autonomous drone swarming technology, from target awareness through launch to termination — an end-to-end offensive system built by a private company. ^[6] And it is seeking private AI data centers on military bases, 207 acres at Fort Hood, with the same eight firms. ^[7] The state sets the targets. The firms build and operate the systems. The legal liability for a lethal AI decision has not been tested in court — the cyber NSPM makes the no-immunity design explicit, but in the AI and drone domains the liability question is still emerging. In drone warfare, the franchise model is most advanced. Ukraine launched online marketplaces — Brave1 and DotChain — where frontline commanders buy drones from hundreds of manufacturers using credits and bonus points, functioning like consumer e-commerce for lethal weapons. ^[8] The US-Ukraine drone export pact brings Ukrainian private firms directly into the American supply chain: UFORCE is partnering with US shipbuilder ReconCraft to manufacture MAGURA autonomous maritime drones in Oregon. ^[9] The Air Force has created a competitive marketplace for autonomous combat algorithms, where private firms — Shield AI, Anduril, General Atomics, Collins Aerospace — compete to provide the software that flies autonomous combat aircraft, while the government sets the standards through a standardized software architecture called A-GRA. ^[10] The state writes the rules. The firms compete to execute within them. Three structural components hold this franchise together, and no single story captures all three. The first is permissive authorization. The state provides legal cover and oversight for willing private firms to conduct operations that would be diplomatically radioactive for the government itself. The August 2026 cyber NSPM is the cleanest example: DOJ and DHS must approve each operation in writing, the state supervises, but the firm — not the government — carries the legal exposure. ^[3] The authorization is real. The immunity is not. The second is coercive compliance. The state does not merely invite private firms to participate. It punishes those that refuse. When Anthropic's CEO Dario Amodei refused to remove the safety guardrails prohibiting Claude's use for mass surveillance and autonomous weapons, the Pentagon designated the company a national security supply chain risk — a label normally reserved for foreign adversaries. ^[11]

America’s warfighters will never be held hostage by the ideological whims of Big Tech. — Pete Hegseth

The Pentagon disputed a $200 million contract with the firm. ^[12] Other contractors were prohibited from working with it. And yet, throughout this public punishment, the US military continued using Claude covertly — for intelligence and targeting during operations in Iran and the operation that captured Nicolás Maduro in Venezuela. ^[13] The state punishes the firm publicly for refusing the franchise and uses its capabilities anyway. Participation is not optional. Safety guardrails are treated as a breach of the franchise agreement. The third is enforcement against unauthorized operators — and this is where the Colombian sanctions come back into view. In December 2025, the US sanctioned A4SI and Maine Global Corp SAS for recruiting fighters for Sudan's RSF paramilitary. ^[14] These were private military actors operating outside US authority, and the state moved to shut them down. Meanwhile, the US was simultaneously incorporating allied private firms — Ukrainian drone manufacturers building in Oregon — as new franchisees. ^[9] The binary the US enforces is stark: operate under its authority and gain contracts and legal cover, or operate outside it and face sanctions. There is a structural aside that sharpens the whole picture. In November 2025, the FCC rescinded Biden-era telecom cybersecurity requirements that had been implemented after the Salt Typhoon Chinese hack — rules that required carriers to secure their networks against foreign intrusion. ^[15] Senator Cantwell made the charge explicit.

You have now proposed to reverse this requirement after heavy lobbying from the very telecommunications carriers whose networks were breached by Chinese hackers. — Maria Cantwell

The state was simultaneously lifting the defensive burden from private firms and granting them offensive authority — less defense required, more offense authorized. The franchise is now being exported as alliance doctrine. NATO is studying Ukraine's decentralized private-sector drone procurement model to modernize Western military capability. ^[16]

The West is now in a race. — Sir John Stringer

Ukrainian drone pilots, operating with commercially procured equipment, defeated Swedish troops in a NATO exercise on Gotland. ^[17] The model is spreading. None of this was inevitable. Russia's parallel shift to drone-centric hybrid warfare — training one million drone operators by 2030 — is state-run, not franchised. ^[18] The European Union went the opposite direction entirely: in February 2026, the European Parliament disabled built-in AI features on official devices.

As these features continue to evolve, the full extent of data shared with service providers is still being assessed. Until this is fully clarified, it is considered safer to keep such features disabled. — European Parliament IT services

The US chose to integrate private AI into the warfighting apparatus while the EU chose to strip it from government devices. The franchise model is a policy choice, not a technological inevitability. What the state keeps, in the end, is what matters: the authority to decide who operates, against whom, and under what rules. What it gives away is the execution, the legal exposure, and the diplomatic deniability that comes with a private actor rather than a government one. That is not a retreat from sovereignty. It is sovereignty exercised through a new instrument — one that can be disowned the moment an operation goes wrong.