“AI is only as good as the data that it gets, and we’re going to make sure that it’s there”, said U.S. Defense Secretary Pete Hegseth in an announcement this week of embedding Elon Musk’s Grok chatbot alongside Google’s Gemini inside Pentagon networks. The move represents one of the most aggressive steps yet in the military’s push to accelerate the adoption of artificial intelligence across both unclassified and classified systems.
It comes amid tightening international scrutiny of Grok-which has come under heavy fire for its use in the creation of non-consensual, sexualized deepfakes since Elon Musk embedded the AI chatbot into his X platform. Malaysia and Indonesia have blocked access to the tool and the UK media regulator opened a formal probe. Hegseth concluded that despite this adverse reaction, Grok would go live inside the DoD systems later this month with access to “all appropriate data” from military IT and intelligence databases.
But embracing Grok is part of a wider shift in the Pentagon’s defense AI strategy. In late 2024, the Biden administration put out a broad national security memorandum on frontier AI that emphasized responsible use, human oversight for critical functions, and bans on such uses as automating nuclear launch decisions. For all its effort to balance speed with safeguarding ethics and operational risks under continuing debate, the approach under Hegseth suggests a willingness to favor rapid deployment.
Hegseth framed his vision as unencumbered by “ideological constraints” that may stand in the way of legitimate military uses-a telling comment on the cultural and political fault lines defining AI policy. Musk has been marketing Grok as a counter to what he describes as “woke AI,” pitting it against competitors including Gemini and ChatGPT. But concerns that Grok has a history of producing antisemitic and extremist content have raised alarms about whether its use in sensitive defense systems can be controlled without producing unwanted outcomes.
Inside the Pentagon, integration is part of a broader AI acceleration strategy. Along with Grok and Gemini, it will launch its own GenAI.mil platform to millions of personnel, offering tools for research, document formatting, and imagery and video analysis. For officials, it is part of the department leveraging decades of operational data for everything from logistics planning to intelligence synthesis. The Chief Digital and Artificial Intelligence Office has been tasked with enforcing data-sharing mandates so AI models have the breadth of information they need to perform effectively.
But the push comes as defense leaders grapple with the security risks of AI-enabled systems. Special Operations Command is testing AI for biometric and communications analysis but weighing the vulnerabilities such tools could introduce if adversaries adopt similar capabilities. A Responsible AI Strategy the Department issued calls trust, safety, and adherence to ethical principles indispensable to sustaining public and allied confidence in military use of AI.
There are also a number of unanswered questions about governance-the Defense Department’s Directive 3000.09 on autonomous weapons requires human judgment in the use of force and requires rigorous testing for any system with autonomous targeting capabilities. Grok is not a weapon, but its integration into mission systems does raise some similar issues about oversight and reliability and compliance with the law of war. A national security AI governance framework called for chief AI officers across agencies to oversee high-impact systems and enforce risk management practices, but their success depends upon consistent execution.
For now, the Pentagon is betting that embedding frontier AI models into its networks will yield decisive advantages in speed, adaptability, and insight into operations. Whether this pays off without eroding safeguards will be a test not only of the technology but of the institutions tasked with controlling it.
