Pentagon Cuts AI Weapons Testing Team Raising Safety Concerns

In a bold and controversial move, the Pentagon has drastically reduced the size of its Office of the Director of Operational Test and Evaluation (DOTE), the key agency responsible for independently testing and validating the safety and effectiveness of weapons and artificial intelligence (AI) systems. This office, established in the 1980s to prevent unproven or unsafe technologies from reaching the battlefield, has been cut from 94 staff members to just 45, with its director replaced and a rapid seven-day implementation deadline.

Secretary of Defense Pete Hegseth framed this overhaul as a necessary step to reduce bureaucracy and waste, promising to save $300 million and speed up the deployment of AI-enabled weapons systems. However, experts warn that this “chainsaw approach” risks undermining the critical safety net that ensures new military technologies perform reliably under real combat conditions.

Missy Cummings, a former US Navy fighter pilot and engineering professor, highlights that DOTE acts as the “last gate before a technology gets to the field,” providing an independent check that helps reduce waste, fraud, and abuse. Without this rigorous testing, there is a heightened risk that AI systems—particularly complex and nascent technologies like large language models—may be deployed with critical flaws or safety issues that only surface in real-world combat scenarios.

This restructuring comes at a pivotal moment as the Pentagon increasingly integrates AI across its operations. Defense companies such as Anduril, which recently secured a $2.5 billion funding round, and AI firms like Anthropic and OpenAI are actively developing military applications. While these companies have close ties to the current administration, they have not publicly endorsed the cuts. Still, the message is clear: obstacles to rapid AI adoption favored by the Trump-era defense approach are being removed.

Mark Cancian, a former Pentagon advisor, expresses concern that while the office’s downsizing may not render testing ineffective, it could allow critical issues to slip through undetected until combat deployment. The unique challenges posed by AI systems—such as unpredictable behaviors and “hallucinations” in generative models—make thorough testing more essential than ever.

Historically, the military has experimented with AI technologies like computer vision for drones, but the current wave involves ambitious pilots deploying large language models and other advanced AI at scale. The reduction of DOTE’s capacity threatens to weaken the final safeguard that ensures these cutting-edge systems are battle-ready and safe for soldiers relying on them.

In essence, the Pentagon’s push for speed and efficiency in AI weapons adoption may come at the expense of reliability and safety. This trade-off raises important questions about how the military balances innovation with risk management in an era where AI systems are becoming increasingly complex and integral to national security.

For defense contractors, this shift opens a faster path to field deployment but also places greater responsibility on companies to ensure their AI technologies meet rigorous standards. For policymakers and military leaders, it underscores the urgent need to develop new frameworks that can effectively evaluate AI systems’ safety and effectiveness without the traditional level of independent oversight.

As AI continues to transform defense capabilities, the question remains: can the Pentagon innovate rapidly while still safeguarding the troops and missions that depend on these technologies? The answer may well shape the future of military AI adoption and national security.