I spent decades inside the Pentagon watching technology reshape warfare. I saw precision munitions change the battlefield. I watched satellites compress decision cycles. But nothing compares to what is happening now.
Artificial intelligence has moved the lab to the kill chain.
And the showdown between Secretary of War Pete Hegseth and AI firm Anthropic is not a contract dispute. It is the opening battle over who controls the most powerful military technology of the 21st century.
AI is already transforming war
Look at Ukraine.
Western officials report that drones now account for roughly 70-80% of battlefield casualties in that war. But the real revolution occurs when AI is added. Reports indicate AI-guided navigation can increase drone strike accuracy from 10–20% to as high as 70–80%.
That is not incremental change. That is a transformation in battlefield lethality.
The same dynamic is emerging in U.S. operations involving Iran and other theaters. AI tools are being used for intelligence analysis, targeting refinement, pattern recognition, and operational simulations. These systems compress time, reduce uncertainty and accelerate decisions.
AI is not theoretical. It is operational.
Which brings us to Washington.
What the Hegseth–Anthropic standoff is really about
On Feb. 27, Hegseth designated Anthropic a ‘supply chain risk to national security.’ President Donald Trump ordered federal agencies to cease using its Claude AI model after Anthropic refused to remove two guardrails:
A prohibition on fully autonomous weapons.
A prohibition on mass domestic surveillance.
Artificial intelligence has moved the lab to the kill chain.
The Pentagon argues that military commanders must be able to use AI tools for all lawful defense purposes without seeking permission from a private company in real time.
Anthropic argues that removing safeguards could enable autonomous killing systems or unconstitutional domestic spying.
Both concerns are legitimate.
But here is the deeper problem: America has outsourced strategic control of its most sensitive military algorithms to private contractors.
That is unsustainable.
Draw the right line
Let me be clear about what must not happen.
We must not expand domestic surveillance of American citizens under the banner of AI efficiency. The Fourth Amendment does not disappear in the age of algorithms.
Second, we must keep a human being in the kill chain. I served under lawful command authority. Life-and-death decisions carry moral accountability. They cannot be delegated entirely to autonomous systems.
Those are firm boundaries.
But here is the other boundary: no private corporation should hold an effective veto over how America defends itself.
Washington’s contractor addiction
For decades, the federal government has grown dependent on contractors for critical defense functions — logistics, cyber infrastructure, analytics and intelligence support. AI is simply the next frontier in that pattern.
But frontier AI models are not spare parts or uniforms. They are strategic infrastructure. They influence targeting, operational tempo and potentially deterrence modeling.
That level of sensitivity cannot remain under corporate ownership.
During World War II, the United States built the atomic bomb through the Manhattan Project under centralized national authority. It was not governed by venture-backed boards setting independent usage policies. It was directed by the U.S. government with a clear strategic mandate.
We need a similar mindset for our most sensitive AI systems.
Government must own core military algorithms. Not lease them. Not subscribe to them. Own them.
AI tools are being used for intelligence analysis, targeting refinement, pattern recognition, and operational simulations. These systems compress time, reduce uncertainty and accelerate decisions.
If AI is the new strategic high ground, America cannot subcontract the high ground.
China isn’t hesitating
As I argue in ‘The New AI Cold War,’ Beijing does not struggle with these dilemmas.
China fuses AI development directly to the state. There are no Silicon Valley executives in Beijing refusing military access. AI is treated as national infrastructure.
Russia and other nations are moving in similar directions. They are not debating internal guardrails while field-testing AI-enabled systems.
Strategic competition does not pause while we litigate contract language.
What must happen next
First, Congress must draw bright lines: no AI-enabled mass domestic surveillance of Americans without strict constitutional safeguards.
Second, codify meaningful human control over lethal force decisions.
Third — and most critically — build sovereign AI capacity inside government.
That means:
- Government-controlled AI research for classified applications
- Government ownership of core defense algorithms
- Reduced reliance on private frontier labs for sensitive military systems
- Long-term pipelines of cleared AI engineers
Anthropic argues that removing safeguards could enable autonomous killing systems or unconstitutional domestic spying.
Private industry will continue to innovate. But America’s most sensitive warfighting tools cannot remain dependent on companies whose corporate policies can override national defense requirements.
The real issue is sovereignty
The Pentagon–Anthropic feud is not about personalities. It is about sovereignty.
Who controls the algorithms that guide American force?
Who owns the code?
Who decides how it is used?
In the new AI Cold War, power will belong to those who control the models — not merely those who rent access to them.
America must protect liberty. We must reject AI-driven domestic surveillance. We must preserve human moral accountability in the use of force.
First, Congress must draw bright lines: no AI-enabled mass domestic surveillance of Americans without strict constitutional safeguards.
But we must also end the illusion that venture-backed firms can function as ultimate gatekeepers of national defense.
The AI Cold War is not hypothetical. It is unfolding on battlefields abroad and in policy fights at home.
This moment is not about one company. It is about whether the United States will treat artificial intelligence as strategic national infrastructure — or as a contractor service.
The answer will shape the next generation of warfare.
And history will not wait for us to decide.