China's AI Arsenal Progressing at 'Disruptive' Pace
China's People's Liberation Army is making steady and sometimes 'disruptive' progress in fielding AI-driven military systems, according to a new Foreign Affairs analysis. The PLA is reportedly seeing significant gains in autonomous drones, electronic warfare, and AI-powered decision-support tools. This progress is a major driver behind the Pentagon's push for faster, dual-use tech adoption.
The Pentagon's drive to counter China is reshaping how tech enters the defense market, with a focus on speed and dual-use capabilities. The Defense Innovation Unit (DIU), for example, aims to move from problem identification to prototype awards in 60-90 days using a competitive Commercial Solutions Opening (CSO) process. Successful prototypes can lead to non-competitive follow-on production contracts, a pathway designed to attract commercial tech firms that traditionally avoid the slower Federal Acquisition Regulation (FAR) system. A key initiative is the DIU's Thunderforge project, which awarded a prototype contract to Scale AI to build an AI-powered decision-making platform for military planning. This system will use advanced large language models and wargaming simulations to accelerate strategic assessments for commands like INDOPACOM and EUCOM. The project underscores a major shift: acquiring enterprise-scale commercial software from companies like Scale AI and Anduril to overhaul core defense processes. To manage this influx of technology, the DoD established the Chief Digital and AI Office (CDAO) and, in partnership with DIU, launched an AI Rapid Capabilities Cell. This cell is tasked with accelerating the adoption of generative AI and has an initial budget of approximately $100 million for pilot programs in areas from command and control to logistics and software development. This creates a clear entry point for companies with mature AI tools looking to engage with the DoD on priority use cases. For smaller contractors, the Small Business Innovation Research (SBIR) program remains a critical gateway. In May 2025, the Army awarded SBIR contracts to Seekr, a commercial AI company, to develop generative AI agents for Project Linchpin, the Army's effort to create a trusted AI/ML operations pipeline. This demonstrates a direct path for startups to integrate their technology into major defense modernization programs. This push for innovation is backed by significant acquisition reform. A January 2025 proposed FAR rule aims to expand small-business set-asides on multiple-award contracts. Furthermore, a May 2025 DoD program called SWIFT (Software Fast-Track) uses AI-driven security assessments to grant provisional Authority-to-Operate (ATO) much faster, tackling a major hurdle for software companies entering the defense sector. As contractors integrate AI into their own workflows for tasks like proposal writing, they must navigate strict security requirements. Handling Controlled Unclassified Information (CUI) requires adherence to standards like DFARS 252.204–7012 and NIST SP 800–171. For cloud-based AI tools, FedRAMP authorization is essential, meaning contractors must use enterprise-grade, secure environments like Azure Government with Azure OpenAI services to process sensitive data. The DoD has also established a clear ethical framework, adopting five principles for Responsible AI (RAI): responsible, equitable, traceable, reliable, and governable. The DIU has embedded these principles into its development process, creating guidelines to ensure fairness, accountability, and transparency are built into AI systems from the start. This focus on governance is a critical consideration for any company looking to become a long-term partner in the defense AI ecosystem.
