WASHINGTON — Chief Technology Officer of the Department of Defense Today, he weighed in on the reported conflict between the Pentagon and AI giant Anthropic, in which he publicly rejected as undemocratic the company’s attempts to limit the military use of Claude AI.
“Congress writes the legislation, the president signs it, government agencies write the regulations, the public complies, and we have always complied,” Undersecretary Emile Michael told reporters after speaking at the Microelectronics Commons Consortium.
“What we’re not going to do is go beyond what Congress passed and have certain companies dictate a set of new policies,” Michael said. said. “That’s not democratic. It’s giving control to one company over what the new policy is, and it’s the president’s, it’s Congress’s, and it’s up to each agency to decide how to implement those rules.”
Last summer, the Department of Defense’s Office of the Chief of Digital and AI awarded contracts worth up to $200 million each to customize popular generative AI applications from Anthropic, Google, xAI, and OpenAI for military use. Semafor reports that a classified version of Anthropic’s Claude AI is also available to Department of Defense personnel through Amazon and Palantir.
However, according to Wal newspaper January reportl Street Journal, Anthropic’s policy prohibiting the use of Claude in weapons and surveillance programs created a rift that jeopardized its contract with the Pentagon. The journal also has at least In one case, Claude was used to help plan an attack that captured Venezuelan strongman Nicolás Maduro.
In the magazine’s report above maduro attack, An Anthropic spokesperson declined to discuss specific operations, but said, “If you use Claude, whether in the private sector or across government, you must follow usage policies that govern how Claude is deployed.”
These policies prohibit using AI to “manufacture, modify, design, or illegally obtain weapons” or “track a person’s physical location, emotional state, or communications without their consent, including the use of our products for battlefield management applications.” (Please note that the prohibition applies to everyone, not just U.S. citizens).
According to reports, the difference of opinion went up to Defense attention secretary pete hegseth Although he is pushing for AI in the Pentagon, he also resents external regulations for the military. An anonymous Pentagon official told Axios that Hegseth was even “close” to designating the company as a “supply chain risk.” It was done It is calling on companies that do business with the Department of Defense (including giants like Microsoft, Google, and Amazon) to sever all ties with Anthropic, including their use of Claude.
The official statement from Pentagon Chief Press Secretary Sean Parnell was more subdued, telling The Hill that “the relationship between the Department of the Army and Anthropic is being reviewed.”
Michael today refrained from making any threats, stressing that current government safeguards are sufficient, but saying he expects Anthropic to succeed.
“This country has a strong set of laws regarding surveillance that have been implemented through a democratic process,” he said. “From an autonomy standpoint, [there] ]There are many regulations that have been promulgated by the department over the years,” he added, addressing questions such as “If a swarm of drones is coming toward a military base, and human reaction time is not fast enough, what are the options for defeating it?”
Anthropic did not immediately respond to Breaking Defense’s request. In response to Michael’s statement.
“I want guardrails” But…
Despite the impasse over usage policy, Michael made it clear that he considers Anthropic one of the “national champions” of AI in the U.S., and said he hopes Anthropic will lift restrictions and continue working with the military, just as Google did after withdrawing from the military’s Project Maven in 2018 due to internal rebellion.
“The great news in the AI field is that the United States is leading the way,” Michael said at the annual meeting of the Microelectronics Commons, a public-private consortium of chipmakers, academics and others working with the Department of Defense. “We have at least four true national champions, probably more, and we are investing $1 trillion in facilities and research and development over the next few years.”
After those remarks, a reporter asked about the future of Anthropic’s contract with the Department of Defense, and Michael quickly turned positive, saying, “The Secretary has said he’s looking at the relationship, so we’re looking at it. We want all of America’s champion AI companies to succeed. I want Anthropic to succeed, xAI, OpenAI, Google to succeed.”
“We want to leverage all the capabilities that we believe will change the world,” he continued. “And if you think back to 2018 when Google didn’t want the Department of the Army to use its cloud business, this is a similar moment.”
Michael added that AI should have appropriate safeguards built in against misuse, even by the Department of Defense. However, the definition of “abuse” must not be so broad as to interfere with legitimate military functions.
“We want guardrails,” he said. “We need guardrails tailored to military applications. We can’t have AI companies sell AI to the Department of the Army, we can’t let them do the War Department’s job, because our job is to protect our country and protect our military.”
