In response to OpenAI's controversial agreement with the U.S. Department of Defense, Caitlin Kalinowski resigned from her position as head of hardware at the company.
“This was not an easy decision. Artificial intelligence plays a crucial role in national security. However, surveillance of Americans without a warrant and lethal autonomy without human approval are issues that deserve more scrutiny than they are currently receiving,” she wrote.
Kalinowski previously worked at Apple on Mac hardware before moving to Meta, where she became a key engineer in the Reality Labs division. She joined OpenAI in 2024 as the head of robotics and consumer hardware.
In her statement, she emphasized that her decision was “about principle, not people,” expressing “deep respect” for OpenAI CEO Sam Altman and the startup's team.
“In my opinion, the announcement was made hastily, without defining boundaries. These are primarily governance issues. They are too important to rush into agreements,” Kalinowski added.
An OpenAI representative stated that the agreement with the Pentagon “creates a pathway for the responsible use of AI in national security while clearly outlining red lines: no internal surveillance and no autonomous weapons.”
“We understand that people have strong beliefs on these issues, and we will continue to discuss them with employees, the government, and society worldwide,” he added.
The agreement between OpenAI and the Pentagon was finalized at the end of February, shortly after negotiations between Anthropic and the Defense Department stalled.
Altman soon acknowledged that the deal was “hasty” and that it “doesn’t look good” for the startup’s image.
This situation has impacted the popularity of ChatGPT and Claude. The number of deletions of the first chatbot surged by 295%, while the latter rose to the top of the App Store rankings.
Mobile app downloads for Claude and ChatGPT in the U.S. Source: Appfigures.
Government's Struggle with Anthropic
In July 2025, the U.S. Department of Defense awarded contracts worth up to $200 million to Anthropic, Google, OpenAI, and xAI for developing AI solutions in security.
In January 2026, the WSJ reported on the risk of a contract termination with Anthropic due to disagreements over the startup's strict ethical policies. These rules prohibit using the Claude model for mass surveillance and autonomous lethal operations.
In February 2026, the U.S. Army utilized Claude in an operation to capture Venezuelan President Nicolás Maduro, escalating the conflict.
CEO Dario Amodei met with Defense Secretary Pete Hegseth to discuss the situation. The department issued an ultimatum: Anthropic must accept government conditions by February 27.
Amodei stated that he would prefer not to cooperate with the Pentagon than agree to use his technologies in ways that could “undermine rather than protect democratic values.”
At the end of February, President Donald Trump ordered all federal agencies to completely cease using Anthropic technologies within six months.
He described the startup as a “runaway” company run by people who “have no idea what the real world is like.”
The Pentagon officially recognized Anthropic as a “supply chain risk,” jeopardizing the firm's business with other government contractors. Microsoft and Google stated they would continue providing products to the company.
Heading to Court
Amodei announced that his company would challenge the Defense Department's decision to label the firm as a supply chain risk in court. He believes it is “legally unfounded.”
The statement indicated that the Pentagon's decision does not affect the overwhelming majority of Claude's clients.
According to Amodei, the company had productive negotiations with the Defense Department for several days, but they fell apart after an internal memo was leaked. In it, the CEO characterized OpenAI's relationship with the department as “more of a show of safety than real risk protection.”
Amodei apologized for the leak, stating that the company unintentionally disclosed the memo's contents and did not instruct anyone to do so.
“It is not in our interest to exacerbate the situation,” he wrote.
The startup head emphasized that the document was written within hours after a series of announcements that Anthropic would be removed from federal systems. He apologized for the tone, calling that day “difficult for the company.”
Amodei added that the memo represents an “outdated assessment” and does not reflect a “thoughtful and measured opinion.” He stated that the company's top priority is to ensure American soldiers and national security experts have access to crucial tools during ongoing major combat operations.
Anthropic will continue to provide the Department of Defense with its models “at cost” for as long as necessary.
Former White House AI advisor Dean Ball noted that the company will find it challenging to contest the Pentagon's decision.
“Courts are quite reluctant to question government decisions on what constitutes a national security issue. A very high bar must be overcome for that. But it is not impossible,” the expert remarked.
It is worth noting that in February, Anthropic softened its key safety policy, citing the need to maintain competitiveness.