Even as governments around the world seek greater control over artificial intelligence, the United States government designated Anthropic as a "supply chain risk" on February 27, 2026. Positions are being taken in the tussle between the executive and corporations over who should regulate AI and how.

Does State sovereignty trump corporate ethical responsibility?

Anthropic was banned from entering into contracts with any US government department on account of their stand against using AI for fully autonomous weapons and domestic surveillance in the US. The government did not stop there; it indicated that this ban would include a secondary ban on any company that has a commercial relationship with Anthropic from entering into contracts with the US government.

This could have far-reaching consequences if the intent expressed translates to legal validity. Most major defence contractors, such as Amazon, Google, and Microsoft, have investments and dealings with Anthropic. What happens to the billions of dollars these companies have invested in Anthropic? Will they be allowed to sell cloud services to Anthropic? Amazon's stock price remained flat after this announcement, suggesting the market does not believe the wider ban will hold but might only prevent them from using Anthropic for specific military contracts with the US government.

While Anthropic was denied a contract and banned, arch-rival OpenAI announced the next day that it had signed a contract with the Pentagon. OpenAI asserts that the contract prohibits using its technology for "mass domestic surveillance or directing fully autonomous weapons and making high-stakes automated decisions." A closer reading of the development suggests that OpenAI's restrictions apply where such use cases are already unlawful, allowing the Pentagon wide discretion. That had, all along, been the position of Peter Hegseth, the US Secretary of War.

The limits of regulation

Corporations believe the executive does not adequately understand the risks and needs their guidance. The executive does not want the corporations to determine what they should and should not do. There is merit in both arguments.

On the one hand, the technology is the private property of the respective corporations, and they should not be forced to sell their technology to the military unconditionally. Corporate leaders such as Dario Amodei argue that they have an ethical duty to prevent their technology from being misused. The US government also has a historical record of exploiting legal loopholes, as it did when it authorised mass collection of civilian phone records, a fact disclosed by Edward Snowden in 2013.

On the other hand, it does seem counter to democratic principles that the ethical commitments of corporate CEOs should restrict governmental action. Decisions about government and military power ought to be made through democratic processes and not by private companies unilaterally. Accepting the corporation's stance in this case would be like an aircraft manufacturer, such as Lockheed Martin, deciding which missions the Pentagon could fly and which ones it could not, according to Alan Z Rozenshtein from the University of Minnesota. This is the case for "supply chain vulnerability".

Regarding the secondary boycott, Rozenshtein opines that the Pentagon cannot force all its contractors to stop using Anthropic or having any commercial relationship with the company. At best, they could prevent the use of Anthropic for executing specific contracts with the Department of Defence. Although the ban will be fully implemented after a six-month transition period, its legality, practicality, and real motive are questionable. There is also a discussion on whether Congress, not the executive, should decide on the ban.