The Pentagon’s AI Dilemma: Navigating Security Risks and Shifting Priorities
The U.S. Department of Defense is grappling with complex challenges in its pursuit of artificial intelligence capabilities, balancing national security concerns with the realities of a rapidly evolving tech landscape. Recent negotiations with AI developers, coupled with personnel choices within the Pentagon, reveal a delicate dance between innovation and risk mitigation.
Pentagon’s Hardball Tactics with Anthropic
The Pentagon is engaged in intense contract renegotiations with Anthropic, designating the company a “supply-chain risk” until it complies with demands regarding its acceptable use policy. This dispute highlights the critical require for control and oversight in the deployment of advanced AI technologies within defense systems. A recent report by The Verge detailed the high stakes involved, noting the Pentagon’s attempt to demonstrate the seriousness of the dispute by including multiple senior officials in meetings with Anthropic CEO Dario Amodei.
Questionable Backgrounds of Key Personnel
The composition of the Pentagon’s negotiating team has raised eyebrows. While some officials have extensive government and military experience, others bring backgrounds marked by controversy:
- Emil Michael: As Pentagon CTO and spearheading negotiations with Anthropic, Michael’s past as second-in-command at Uber during Travis Kalanick’s tenure is under scrutiny. He was ousted from Uber in 2017 following an investigation into a culture of sexual harassment and allegations that he suggested hiring opposition researchers to discredit critical journalists. Design Thinking: The Answer to the Impasse Between Innovation references Uber’s extensive lobbying efforts and the controversies surrounding its practices.
- Steve Feinberg: The Deputy Secretary, founder of Cerberus Capital Management, has faced criticism for his role in the demise of Chrysler and his early support for Donald Trump. His firm, Cerberus, has significant investments in defense companies, raising potential conflicts of interest. Cerberus Ventures, launched in 2023, focuses on venture capital investments in national security-related technologies.
- Sean Parnell: Hegseth’s chief spokesperson, Parnell, previously ran for a Senate seat in Pennsylvania but withdrew after allegations of abuse surfaced during a custody hearing.
These appointments underscore a trend of bringing private sector mindsets into government negotiations, potentially prioritizing business interests over national security concerns.
The Single-Supplier Vulnerability
A key issue driving the Pentagon’s urgency is the “single-supplier vulnerability” related to AI. A 2024 national security memorandum mandated that the Department of Defense maintain contracts with at least two frontier AI labs cleared for classified information to avoid reliance on a single vendor. However, as of early 2025, only Anthropic’s model was cleared for classified use. This situation forced the Pentagon to grant xAI’s Grok access to classified systems, despite acknowledging that Grok is less advanced and reliable than Anthropic’s Claude.
A Limited Field of Choices
The Pentagon’s options are constrained, presenting a difficult trade-off:
- Anthropic: A company with a strong AI model but increasingly flexible ethical boundaries.
- A model refusing deployment for autonomous lethal applications.
- Models not yet secure enough for classified networks.
- xAI’s Grok: Considered less advanced and reliable.
As one Defense official told Axios, “The only reason we’re still talking to these people [Anthropic] is we need them and we need them now. The problem for these guys is they are that fine.”
Political Influences and External Factors
The situation is further complicated by political influences. The Regulator newsletter highlights the involvement of figures like Laura Loomer, who attempts to influence policy debates through social media campaigns, and the unexpected financial connections within Trumpworld, such as the purchase of a rare Pokémon card by the son of former Trump ally Anthony Scaramucci.
Legal Risks and AI Platforms
Recent legal developments underscore the importance of protecting privileged information when using AI platforms. The Southern District of New York has cautioned against entering attorney-client correspondence into publicly available AI systems, as it can waive attorney-client privilege.
The Pentagon’s pursuit of AI capabilities is a complex undertaking fraught with technical, ethical, and political challenges. Navigating these challenges will require careful consideration of security risks, responsible personnel choices, and a commitment to maintaining a diverse and resilient AI supply chain.
Worth a look