While We Were Bombing

When it comes to the fight between Anthropic and the Pentagon, I’m on Team Claude. If asked to trust Anthropic CEO Dario Amodei or Secretary Pete Hegseth, I’m picking Dr. Amodei. The spat between Anthropic and the Pentagon may really be less about AI governance than a personality problem between the two men, but still is important. All that being said, I hate to break the news to Dr. Amodei, but there are going to be autonomous AI weapons – if there are not already – and AI is almost certainly already being used for mass surveillance, even of U.S. citizens.

Attack of the drones -- guided by AI. Credit: Microsoft Designer

Those were his supposed “red lines,” and they are good ones, but technology advances and current events have rendered them moot. Are they “lawful”? Well, they probably aren’t illegal, but that speaks more to how outdated our laws are when it comes to AI (or many other newer technologies). Meanwhile, of course, the U.S. and Israel unilaterally attacked Iran – pick your choice of the many rationales offered – and Claude has been an integral part.

The future of war has arrived. It actually arrived in Ukraine a couple of years ago. A war that started out as a 20^th century war, relying heavily on tanks, troops, and artillery, quickly evolved into something few had been expecting -- a war of drones, cell phones, GPS, AI, anti-drone countermeasures. Ukraine has demonstrated startling (and desperately needed) innovation, in tactics, strategy, and especially drones. Despite the country being battered by Russian missile and drone attacks, Ukraine produces over 4 million drones a year, far more than the U.S. or, indeed, all NATO countries combined.

U.S.-supplied missile systems like the Patriot, Stinger, or Javelins have helped Ukraine fend off Russian attacks, but those systems are expensive and in short supply. And once Russia started using Iranian-designed drones in mass attacks, they became woefully inadequate, not to mention not cost-effective – a $1,000 drone versus a $1 million interceptor?  The economics are clear.

Iranian Shahed drones. Credit: AP

They may be clear, but evidently not quite fully apparent to the U.S. military. Attacks on Iran look a lot like the Gulf War, although the aircraft and the “smart” munitions are better (and more expensive). When Iran retaliated, it was largely through its vaunted Shehed drones. The initial U.S. casualties were the result of a drone attack, as were attacks on U.S. radar systems. As Ukraine painfully learned, but the U.S. apparently did not, expensive missile systems are not well designed to counter massive drone attacks. The Hill reports that Pentagon officials admitted to Congressional leaders that Iranian drone attacks were getting through U.S. defenses, putting our troops and bases at risk.

Kelly Grieco, a senior fellow at the Stimson Center think tank, told The Hill: “It’s worth saying that the notion the U.S. military couldn’t have predicted this threat begs belief given that it was well known about Iranians’ Shahed threat. And we’ve had four years of watching Ukraine deal with Iranian drones and Russian-made variants of them in attacks, so this shouldn’t have come as a surprise.”

And yet…

It’s no wonder that, after years of having to beg for U.S. support, Ukraine’s President Zelensky has offered to share some of his country’s hard won drone expertise. “Our military possesses the necessary capabilities,” President Zelensky said in a post on X. “Ukrainian experts will operate on-site, and teams are already coordinating these efforts.”

Let us hope that U.S. and Israeli officials are not too proud, or too stupid, to take such assistance.

The problem may boil down to, as The Pentagon’s first AI chief, retired Air Force Lt. Gen. Jack Shanahan, told The Wall Street Journal: “The Department of Defense was built as a hardware company in the industrial age, and it has struggled to become a digital company in a software-centric era,” Weapons and weapons systems that take years to develop, more years to produce, while costing cost tens of millions or more, are going to struggle to keep up in a world where weapons can be 3D printed and guided by AI.

It should be noted that last fall President Zelensky warned the UN: “Dear leaders, we are now living through the most destructive arms race in human history because this time, it includes artificial intelligence. We need global rules now for how AI can be used in weapons. And this is just as urgent as preventing the spread of nuclear weapons.” Dr. Dario was perhaps listening, with Secretary Hegseth almost certainly was not.

Craig Jones, a political geographer at Newcastle University, UK, told Nicola Jones in Nature: “The current failure to regulate AI warfare, or to pause its usage until there is some agreement on lawful usage, seems to suggest potential proliferation of AI warfare is imminent.”

It’s here. Is AI being used in Iran? You bet, as Michael Daniels and Dov Lieber of The Wall Street Journal outline, everything from logistics and intelligence analysis to targeting.  

Unfortunately, as Steve Feldstein, a senior fellow at the Carnegie Endowment for International Peace think tank, told Rest of World: “Do we have the right rules in place and accountability norms to handle the exponential growing use of these tools? My answer would be no.”  

Similarly, Daniel Castro, a vice president at Information Technology and Innovation Foundation (ITIF), wrote in IEEE Spectrum regarding the Amodei/Hegseth dispute:

Reasonable people can disagree about where those lines should be drawn

But that disagreement underscores a deeper point: the boundaries of military AI use should not be settled through ad hoc negotiations between a Cabinet secretary and a CEO. Nor should they be determined by which side can exert greater contractual leverage.

If the U.S. government believes certain AI capabilities are essential to national defense, that position should be articulated openly. It should be debated in Congress, and reflected in doctrine, oversight mechanisms and statutory frameworks. The rules should be clear — not only to companies, but to the public.

The Pentagon’s strategy seems to be bombs away, even if it costs $1 billion per day and soon will have diminishing impact. The Administration’s AI strategy seems to be that guardrails only would dampen innovation and leave us behind in the AI race. Our septuagenarian Congress still can’t figure out Facebook, and wants no part of tackling AI. None of these inspire confidence.

I think Dr. Amodei and President Zelensky have a much better grasp on the future – which is already happening -- than do Secretary Hegseth or President Trump, but I worry we’re going to have to go through a lot of scary things before we settle into that future.