Select Language:
OpenAI announced on Friday that it has reached an agreement to allow the Pentagon to access its AI models within the Department of Defense’s classified network, with established safeguards. This comes after former President Donald Trump banned AI competitor Anthropic from government use, citing national security concerns due to its refusal to allow unconditional military deployment of its Claude models. Anthropic has vowed to file a lawsuit, criticizing what it describes as intimidation tactics, and maintains that its technology should not be used for mass surveillance or autonomous weapons.
Later that day, OpenAI CEO Sam Altman revealed a deal with the Pentagon to utilize its models under similar restrictions to those imposed on Anthropic, incorporating “technical safeguards” approved by the Department of Defense. Altman emphasized that two core safety principles—preventing domestic mass surveillance and ensuring human oversight in the use of force, including autonomous weapons—are integral to the agreement.
The Department of Defense has yet to comment publicly. Meanwhile, the U.S. government has publicly rebuked Anthropic, asserting that the military operates within the law and that contractors cannot dictate how their products are utilized. Trump publicly ordered all federal agencies to cease use of Anthropic’s technology, stating, “We don’t need it, we don’t want it, and won’t do business with them again.” He also warned that any non-compliance could lead to severe civil and criminal consequences.
Altman told employees Thursday that he was actively working to negotiate with the Pentagon for an agreement reflecting similar demands to Anthropic’s, hoping to facilitate a resolution. He underscored that human oversight should remain in place for high-stakes automated decisions. Anthropic reaffirmed its stance, asserting that no political pressure would alter its opposition to domestic surveillance and autonomous weapons systems, and expressed readiness to continue supporting U.S. national security.
The Pentagon had set a deadline of 5:01 pm (22:01 GMT) on Friday for Anthropic to comply or face enforcement under the Defense Production Act, a broad authority last used during the COVID-19 pandemic to direct private industry toward national security goals. The department also indicated it might classify Anthropic as a supply chain risk—typically a designation reserved for companies from other nations considered adversarial. In response, Anthropic announced plans to challenge the designation in court, warning of the risks posed by Pentagon demands.
U.S. Secretary of Defense Pete Hegseth directed the Pentagon to proceed with the threat of banning any business activities with Anthropic, criticizing the company for what he characterized as arrogance and betrayal. Congresswoman Hakeem Jeffries, a leading Democrat, praised Anthropic for standing firm against what he called an invasion of privacy and mass surveillance.
The industry has shown solidarity, with hundreds of employees from Google, DeepMind, and OpenAI encouraging their companies to support Anthropic’s stance through an open letter titled “We Will Not Be Divided.” The letter urged leaders to reject government demands for surveillance and autonomous lethal systems, warning that division among companies only benefits those seeking to exert control over AI development.
ChatGPT
Perplexity
Gemini AI
Grok AI
