The U.S. Pentagon is increasing pressure on Anthropic to provide AI without ethical restrictions. Anthropic does not seem to intend to back down, despite government threats.
The Pentagon and U.S. Defense Secretary Pete Hegseth are intensifying pressure on Anthropic. The AI company must drop its restrictions on the use of AI by Friday or face the consequences. Those consequences are significant: if Hegseth does not get his way, he intends to label Anthropic a ‘supply chain risk.’
That label, normally intended for companies with ties to hostile foreign powers, would hit Anthropic economically. After all, neither the Pentagon nor companies doing business with the Pentagon are allowed to work with such a high-risk company.
Terms of Service
The dispute centers on the terms of service linked to Anthropic’s AI model, Claude. The company is willing to be flexible regarding those terms for use within the U.S. military, but it maintains certain principles. For instance, Anthropic wants to prevent Claude from being used for large-scale spying on American citizens or for developing weapons that can fire autonomously (without some form of human intervention).
Hegseth believes it is not up to Anthropic to impose such conditions and is choosing an aggressive approach to force the company to back down. We previously described how this unprecedented government intimidation regarding a company’s internal ethical guidelines could be detrimental to the development of ethical AI in general.
DPA
The Secretary is now also threatening to invoke the Defense Production Act (DPA). This allows the U.S., in the name of national security, to compel a company to develop a specific custom product.
The American business climate is changing due to government pressure: those who hold a different political opinion than the current administration can expect economic penalties.
The Pentagon has essentially backed itself into a corner. At the end of his term, President Biden drafted a memorandum stating that the Department of Defense must not tie itself to a single AI system and must partner with multiple manufacturers. This has not happened under President Trump: Anthropic’s Claude is currently the only system approved for use within the U.S. military. In other words, the Pentagon cannot simply show Anthropic the door.
No intention to back down
Despite the pressure, Anthropic is not yielding. Reuters reports, citing anonymous sources, that CEO Dario Amodei has no plans to lift restrictions on AI for domestic mass surveillance of citizens and autonomous killer robots.
The deadline Hegseth imposed on Anthropic expires Friday afternoon. If Amodei stands his ground, the government could either invoke the DPA or label Anthropic a high-risk company. Both actions are unprecedented in the context of a dispute over terms of service.