The AI Race Won’t Wait for Congress
By Sarah Mathew
Over the first few weeks of March, as the Anthropic and Pentagon saga unfolded at the federal level, one thing became clear: None of this had to happen.
Anthropic, positioning itself as the “AI lab with the greatest emphasis on safety,” is currently in a standoff with the U.S. Department of Defense. Earlier this year, Defense Secretary Pete Hegseth and Anthropic CEO Dario Amodei met to discuss their ongoing collaboration, which began in July 2025 with the goal of Anthropic “prototyping frontier AI capabilities for national security.”
But the relationship quickly soured. When Anthropic allegedly refused to give the Pentagon unfettered access to its large language model Claude, the U.S. government labeled the AI company a “supply chain risk”. Now, a couple of weeks later, Anthropic is suing the Pentagon.
This battle is not just about an AI company. It highlights a governance gap in our system that cuts the United States’ competitive edge, security, and credibility at a critical point in the AI race. Proper governance could have mitigated this fracturing between the private sector and the government.
In its negotiations with the Pentagon, Anthropic stood firm on two red lines: Its product could not be used to launch autonomous weapons and its technology could not be deployed for unconstitutional surveillance of the American people. In response, the Pentagon stated that it would use the AI systems for “all lawful use,” with no exceptions. Secretary Hegseth then invoked historical statutes on Anthropic that had been created for foreign adversaries—the closest available tool in the absence of a federal AI framework. As a result, the contract dispute escalated to the courts because Anthropic felt that its "reputation and core First Amendment freedoms were under attack." The Pentagon’s desire to deploy cutting-edge AI for national security is not unreasonable. The problem is that no agreed-upon framework exists yet to separate overreach from legal use, even when the government is the user.
Home to almost every major AI company, the United States has yet to enact a federal AI law, create an oversight body, or build a framework for military use. Executive orders have served as a starting point, but they cannot substitute for law. In an era of rapid development and deployment, strong AI policy is not only good governance; it is a national security imperative.
The Anthropic-Pentagon duel brings a key unanswered question to public attention: Whose right is it to govern AI?
The answer is straightforward. Private companies build, while elected officials govern. This public-private division exists because although companies initiate technical progress, they ultimately answer to their boards. Congress, on the other hand, is accountable to the American people. In 1958, Congress established the Federal Aviation Administration because public safety demanded it. The same principle should apply to national security. Congress trusted the aircraft industry to innovate, but they knew it was their job to implement safety standards. AI is no different.
Critics argue that private sector innovation moves faster than Congress can regulate, which risks slowing companies down. But in the case of Anthropic, the opposite is true. The absence of a strong framework fuels chaos: marked by a failed contract, a federal lawsuit, and uncertainty at a pivotal time for national security. As Anthropic is stuck in courtrooms for refusing to yield to political pressure, its competitors charge forward. OpenAI continues to scale. China races to innovate. We can debate what an AI framework should look like. What we cannot debate is that the absence of good policy creates a chilling effect, punishing dissent and rewarding competitors who are willing to operate without limits.
No business or government should have unchecked power. Congress must pass binding AI legislation that defines prohibited uses, protects children, creates a clear knowledge-transfer framework, establishes a lean coordinating committee, and ensures private companies can contribute to the process without controlling it. Republican Sen. Marsha Blackburn (R-Tennessee)’s proposed Trump AI Act, drawing on ideas from proposals across the aisle, offers a promising foundation. Congress must maintain the momentum. The goal is not to hinder innovation, but to ensure it serves the American people and protects national security. The framework does not exist yet. That needs to change now.