Same question. Two responses.
One you can defend.

The AI Act explicitly excludes AI systems used solely for military, defense, or national security purposes from its scope.

However, many large U.S. defense primes (Lockheed, RTX, Northrop, General Dynamics, Boeing, etc.) operate substantial civilian or dual-use businesses in aerospace, space, cyber, logistics, and security, which will fall squarely under the Act whenever their AI systems are placed on the EU market or affect EU residents.

The Act operates extraterritorially: any provider or deployer, regardless of location, is covered if they market AI systems in the EU or if the system’s outputs are used there, which clearly captures U.S. firms’ commercial and many dual-use offerings.

Illustration: A U.S. defense contractor selling AI-enabled maintenance and predictive analytics for civilian aircraft fleets in Europe must comply with EU AI Act “high-risk” rules, even though its purely military targeting software does not fall under the Act directly.

For contractors with any EU-facing business, the Act essentially imposes a Brussels-effect compliance layer on their AI engineering and governance.

High-risk classification: AI used in critical infrastructure, transport, border management, biometric ID, and employment decisions is presumptively “high-risk,” triggering stringent obligations for data quality, documentation, human oversight, robustness, and conformity assessments. This likely hits AI-enabled air traffic management, predictive maintenance for civil aircraft, border surveillance and security analytics, and HR/security-vetting tools.

Compliance costs: U.S. companies face requirements to implement risk management frameworks, detailed technical documentation, logging, post-market monitoring, and governance structures. This means creating AI assurance teams parallel to export-control and ITAR compliance, aligning software development lifecycles with EU requirements, and potentially segmenting product lines into “EU-compliant” and “rest-of-world.”

Liability and market-access risk: Penalties can reach up to €35 million or 7% of global annual turnover for prohibited practices. For large primes with tens of billions in revenue, this shifts AI compliance from a legal nuance to a board-level risk.

Brussels Effect pressure: Because it is costly to maintain different AI governance regimes by region, many multinationals are expected to adopt EU-style safeguards globally for high-risk AI.

Although the Act excludes strictly military AI, the institutional and technical practices it requires will bleed into defense R&D. Civil–military cross-pollination means U.S. primes working on dual-use systems are likely to reuse common models, data pipelines, and documentation frameworks across portfolios.

The Act’s emphasis on human oversight and bans on certain manipulative or social-scoring systems will feed into European—and to some extent transatlantic—debates over lethal autonomous weapons and battlefield decision support.

As EU AI safety and transparency requirements harden, Washington may insist that certain defense-relevant AI developed under EU AI Act constraints not be exported to adversaries.

Regulatory divergence as friction: The EU has opted for a comprehensive, binding framework with extraterritorial reach, whereas U.S. policy remains more fragmented and sectoral. This can increase compliance costs and trigger WTO-style debates over technical barriers to trade.

Coordination platform: The EU–U.S. Trade and Technology Council (TTC) is charged with aligning approaches to AI risk management. Washington has an interest in mutual recognition arrangements and interoperability of conformity assessments.

Standards competition: Policy commentary frames the AI Act as setting a de facto global benchmark for AI governance, much as GDPR did for privacy. Third countries may copy or approximate the EU regime.

From an international-political-economy perspective, the AI Act’s main significance is as a structural constraint and bargaining chip in transatlantic tech relations, rather than as a direct restriction on U.S. military AI.

First-mover advantage goes to firms that build credible, documentable AI safety and governance regimes across their civil and dual-use portfolios. There is a real risk of fragmentation if some primes decide EU compliance is too burdensome. Over time, the AI Act will become a reference point in negotiations over export-controlled AI, joint R&D, and arms sales.