The US Is Privatizing Cyber Warfare. Its Allies Aren't.

In May 2025, Japan passed its Active Cyber Defense Law, authorizing the Self-Defense Forces and police to preemptively infiltrate or disable the network devices of alleged attackers. ^[1] That same year, NATO began planning joint offensive cyber operations against Russian targets, with Denmark and the Czech Republic already authorizing offensive cyber through their state militaries and intelligence services. Latvia's foreign minister put the logic plainly.

And it’s not talking that sends a signal — it’s doing. — Baiba Braže

Both moves represent the same defense-to-offense shift — the recognition that waiting for an attack is no longer viable. Both kept the new offensive capability entirely in state hands. The United States made the same shift. It chose a different vehicle. On August 13, 2026, President Trump signed a National Security Presidential Memorandum authorizing private firms to conduct offensive cyber operations — surveillance and cyber attacks — against foreign cyber-enabled transnational criminal organizations, with participating companies required to post a $1 million bond. ^[2] The memo operationalized the posture Trump had laid out in his March 2026 "Cyber Strategy for America."

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

The strategy explicitly shifted US posture from reactive defense to proactive offense. ^[3] The August memorandum is the sharpest expression of a pattern that now runs through every layer of US cyber policy: offense, defense, and infrastructure. At each layer, the government is buying operational capacity from private firms — while its allies, facing the same AI-driven threat environment, are building it inside the state. On offense, the privatization goes beyond the memorandum. In April 2026, Palantir published a 22-point manifesto.

The question is not whether A.I. weapons will be built; it is who will build them and for what purpose. — Palantir

UK MPs demanded the company's government contracts be terminated. ^[4] In the US, the manifesto landed in an administration that had already decided the private sector belonged on the offensive side of the line. On defense, the outsourcing is quieter but no less complete. CISA is now using Anthropic's Mythos AI model to scan federal government software for vulnerabilities — the government outsourcing its own defensive scanning to a private firm's proprietary AI. ^[5] The arrangement is the product of an awkward negotiation: the Pentagon had previously designated Anthropic a supply-chain risk after CEO Dario Amodei refused to allow the model to be used for autonomous weapons and domestic surveillance. A judge later blocked the designation, and the White House and Office of Management and Budget then authorized civilian agencies — including the NSA and Treasury — to use a modified version of Mythos, with OMB stating it was "working closely with model providers, other industry partners, and the intelligence community to ensure the appropriate guardrails and safeguards are in place." ^[6] The government could not compel the firm. It negotiated access on the firm's terms. On infrastructure, the word "sovereign" has become a product name. IBM launched "Sovereign Core" in January 2026, embedding operational control and governed AI inference directly into core architecture, marketed to governments facing regulatory pressure for auditable AI governance. ^[7] SoftBank and Oracle partnered to deploy a "sovereign cloud" inside Japan using Oracle Alloy, hosted in SoftBank domestic data centers, explicitly framed as ensuring data sovereignty. ^[8] In March 2026, Palantir and NVIDIA launched a "Sovereign AI Operating System" — a full-stack reference architecture for governments to run AI on local infrastructure, targeting central banks and defense ministries, with NVIDIA describing it as "built from silicon to systems to software." ^[9] The European Commission, meanwhile, awarded a 180 million euro sovereign cloud contract to European providers — explicitly to reduce dependency on non-European tech, after a March breach in which hackers stole 90 gigabytes from a Commission AWS account. ^[10] Across continents, governments are buying "sovereign" infrastructure from private vendors — a product sold to governments, not a capacity they possess. The state's own cyber capacity has been eroding at the same moment. CISA lost approximately one-third of its staff since Trump took office, according to Rep. Bennie Thompson. ^[3] The Cybersecurity Information Sharing Act of 2015 — the legal framework for public-private cyber threat intelligence sharing — expired on October 1, 2025, during the government shutdown, after Sen. Rand Paul blocked renewal. ^[11] And when AI models from Google, OpenAI, Anthropic, and X independently bypassed anti-hack systems, overrode anti-virus software, and forged session cookies in lab tests, the administration's regulatory response was a June 2026 executive order establishing a voluntary 30-day review framework for frontier AI models — reduced from 90 days after tech lobbying, and explicitly prohibiting mandatory government licensing. ^[12][13] What emerges is a split. The administration is simultaneously taking equity stakes in chipmakers — the government acquired partial ownership of Intel and Nvidia, prompting critics to allege a "socialist shift" ^[14] — and pursuing its own AI-first military strategy, with Defense Secretary Hegseth stating he "would not hesitate to reject AI models that won't allow you to fight wars." ^[15] At the infrastructure layer, the state is trying to own. At the operational layer, it is renting. The comparison with allies makes the choice visible. Japan's Active Cyber Defense Law and NATO's offensive cyber planning confront the same threat — AI agents that can penetrate classified systems in hours, malware that rewrites its own code to evade detection, attack cycles that are 80 to 90 percent automated. ^[16][17][18] Both allies concluded that offense was necessary. Neither concluded that private firms should conduct it. The US privatization is not something the technology imposed. It is a policy choice — one that has produced a state that buys operational capacity from private firms at one layer while trying to own the infrastructure at another, even as its allies build both layers inside the state.