Ahead of the U.S. and Israeli attack on Iran over the weekend, the Pentagon was in tense negotiations with artificial intelligence (AI) company Anthropic over how the Pentagon could use the company’s technology.
Anthropic guarantees that its Claude system will not be used for purposes such as operating autonomous weapons without oversight or human control within the United States.
In response, US President Donald Trump on Friday ordered all US federal agencies to stop using Anthropic’s technology, saying: “We will never allow the woke corporations of the radical left to determine how our great military fights and wins wars!”
Hours later, rival AI research institute OpenAI (maker of ChatGPT) announced it had signed its own contract with the Department of Defense. The main difference appears to be that OpenAI allows “all lawful uses” of its tools, without specifying any ethical lines that OpenAI will not cross.
What does this mean for military AI? Is this the end of the idea of ”ethical AI” in warfare?
AI companies and regulations
Last week’s events come at an already worrying time for AI ethics. Last year, the Trump administration banned states from regulating AI, saying it threatened innovation.
Meanwhile, many AI companies are aligning themselves with the administration, with executives including OpenAI head Sam Altman contributing millions of dollars to President Trump’s inaugural fund. (Mr. Altman said at the time that he also donated to Democratic politicians.)
Anthrop has been less vocal, warning that while working on national security, AI could sometimes undermine democracy and that current systems are not reliable enough to power fully autonomous weapons.
A new international consensus
Much of the concern about military applications of AI focuses on lethal autonomous weapons systems. These are devices and software that can select and attack targets without human intervention.
Just a few years ago, an international consensus seemed to be emerging among governments and technology companies about the dangers of these weapons.
In February 2020, the U.S. Department of Defense released principles for the use of AI across the organization. AI needs to be responsible, fair, traceable, trustworthy, and governable.
Similarly, NATO developed similar principles in 2021, and the UK in 2022.
The United States plays a unique leadership role among its international allies in shaping global norms regarding military conduct. These principles signaled to countries such as Russia, China, Brazil, and India how they believe the United States and its allies should govern the military use of AI.
Military AI and private enterprise
Military AI has relied heavily on partnerships with private industry, as cutting-edge technologies have been developed by private companies.
Project Maven, launched in 2017 with the goal of increasing the use of machine learning and data integration in U.S. military intelligence, relied heavily on commercial technology companies.
The U.S. Defense Innovation Board noted in 2019 that all of the critical data, knowledge, and talent for AI resides in the private sector.
This is still true today. However, norms around how AI is used are rapidly changing in both government and industry.
Trump and Silicon Valley
When Trump was re-elected in 2024, many in Silicon Valley welcomed the prospect of less regulation. Billionaire venture capitalist Marc Andreessen, author of The Techno-Optimist Manifesto, claimed that Trump’s victory “felt like a sore throat.”
Joe Lonsdale, co-founder of AI-based data analytics company Palantir, is another vocal supporter of Trump. Greg Brockman, president and co-founder of OpenAI, personally donated $25 million to pro-Trump organizations last year.
We are a long way from the days of 2019 and 2020.
AI ethics presupposes democratic norms
The question of whether AI-enabled systems are ethical is often seen as a question about the technology itself, rather than how it is used.
In this view, with proper design, AI systems can be created that are inherently ethical. This often involves “algorithmic transparency,” or being clear and honest about the rules the system uses to make decisions. The idea here is that ethics can be “built into” these rules.
The idea of ethical military AI also assumes that it operates on democratic principles. The idea behind algorithmic transparency is that in a democracy, the people ultimately hold power, so they need to know how these systems work.
But in a dictatorship, it doesn’t matter how transparent the algorithm is. There is no sense that private citizens have a stake, no sense that they have a right to know what the government is doing and whether its activities are in accordance with the law.
Free and open debate is considered an important feature of a liberal democracy. While final consensus may be valued, constructive disagreement and even conflict can be a sign of a healthy democracy.
decisions and results
Seen from this perspective, anthropics’ desire to seriously discuss ethical red lines with governments is an example of democracy in action. The company demonstrated both a desire for rational communication and the value of constructive disagreement.
In return, the Trump administration on Friday labeled Anthropic a “supply chain risk,” a rare designation previously reserved for foreign companies, with Defense Secretary Pete Hegseth writing:
Effective immediately, contractors, suppliers, or partners who do business with the U.S. military may not engage in any commercial activity with Anthropic.
Anthropic plans to challenge this declaration in court, as it could have a significant financial and reputational impact on the company.
OpenAI, on the other hand, largely acknowledges that there are no ethical limits, only legal limits. As a result, the company has been able to do business with the U.S. government, but it faces its own reputational consequences as consumer backlash grows.
AI in a world without democratic norms
Ethics in the military What does it mean for AI? One conclusion that is difficult to avoid is that if we want military AI to be used in an ethical manner (according to transparent rules and laws), we need strong democratic norms, which are at risk as the rules-based international order collapses.
So far, little has actually changed. Just hours after President Trump’s condemnation of Anthropic, the United States launched an attack on Iran. The attack was reportedly planned using the company’s software.
