Featured Column: Tony Napolitano: Sovereign AI Abroad, Constitutional Liberty at Home
Opponents of the Pentagon’s desire for sovereign AI, allowing military decision-making unchained from Silicon Valley’s corporate ethics, have spiked the debate with a poison pill that any serious discussion must distinguish: domestic surveillance. The trouble is that the federal government’s domestic authority and its authority to act on a foreign battlefield are distinct from one another both in Constitutional law and in practical reality. To conflate these two arenas of action does harm to both and stifles not only the debate we need but the opportunity to provide nuanced solutions for each.
The federal government was created to wage war, repel invasion, conduct foreign affairs, and provide for the common defense. At the same time, the Constitution deliberately fragmented government power once it turned inward toward the American people. The Bill of Rights was not written to constrain battlefield conduct against foreign adversaries. It was written to constrain the government’s conduct toward its own citizens.
As artificial intelligence rapidly transforms military and intelligence systems, many commentators are collapsing two fundamentally different questions into one:
- Should the United States Department of War develop and control advanced AI warfighting technology to use against foreign enemies, unrestricted by private corporations and their interests?
- Should the United States Department of War be permitted to monitor, classify, predict, or suppress the conduct of Americans domestically?
Those are not the same issue. Treating them as interchangeable is intellectually sloppy and politically dangerous.
The answer to the first question is “yes.”
Our Constitution requires civilian control of the U.S. military, but the corporate leaders of technology companies are not it. Correctly, the Constitution spreads out control of the military by appointing the elected President its Commander-in-Chief and vesting in the elected Congress the power to declare war, raise an army, maintain a navy, and “make Rules for the Government and Regulation of the land and naval Forces.” No corporate board or group of private executives is empowered to override military decisions, regardless of whether the weapons system involved is a chat bot or a boot knife.
To hold otherwise places the power to control national security in the hands of a small cadre of unelected, unaccountable elites. It matters not what the job title or alleged motives of those elites may be–that group is not the proper repository for such power. American servicemembers’ lives are at stake on the battlefield, and they should not be denied a tool that could give them an upper hand or a better chance of making it home alive because its use offends some corporate sensibility. And while big tech companies want to frame themselves as benevolent toward all mankind, giving them the power to deny the use of an AI weapon to those servicemembers may be motivated by profit just as much as anything else.
This is why companies that prioritize sovereign AI, where the customer retains full control, offer significant promise. For example, companies like Cohere have value proposition that customers own the model, own the data, and retain full operational control and decision-making authority. That may be attractive in the private sector, but for the federal government it is essential and should be a baseline requirement. Chuck Flint, the executive director of the Coalition for Affordability and Prosperity, recently made the case in the Washington Examiner that sovereign AI is the “next economic battleground.” And he’s right.
As to the second question: No. The United States Department of War should not be permitted to monitor, classify, predict, or suppress the conduct of Americans, especially not on American soil. Doing so is not its legal mission. When it comes to the comprehensive domestic surveillance discussed in the sovereign AI debate, no part of the federal government has the authority for such a program. AI-powered or not. Not the FBI, not the EPA, not the Bureau of Land Management or Department of Housing and Urban Development.
But who decides which government action is permitted or not? We the People.
This right to be free of such unlawful searches and seizures is a Natural Right enshrined in our Constitution and bolstered by our system of laws. It neither arises from nor is granted by the great philanthropy and restraint of a technocratic cabal. The Constitution and laws of the United States must be followed. The specific language at the heart of the debate is “any lawful use,” and great emphasis should be put on the term lawful. The rhetorical marriage of warfighting control and domestic surveillance in the sovereign AI debate is a red herring.
The Founders made clear that the government is not to be trusted, but neither is any particular faction among the people. If we need greater protections, our elected representatives have a duty to enact legislation safeguarding our rights, and the executive branch must comply. This is why we have judicial review, legislative oversight, and, ultimately, elections.
What should alarm Americans is the emerging belief that private corporations may independently decide which national-security tools the government is morally permitted to use against foreign adversaries.
That is not constitutional governance but the outsourcing of our national interest.
