Caitlin Kalinowski, the leader of the robotics team at OpenAI, has announced her resignation, pointing to the company’s decision to use its artificial intelligence (AI) models inside the Pentagon’s secret network.

“In national security, AI has a crucial role to play. The choice to quit was a tough one, according to Kalinowski, but the boundaries of American surveillance without judicial supervision and deadly autonomy without human permission warrant further consideration.

In an email statement, OpenAI verified Kalinowski’s departure, stating that it feels the Defense Department deal “creates a workable path for responsible national security uses of AI while making clear our red lines, no domestic surveillance and no autonomous weapons.”

“We recognize that individuals have strong opinions on these subjects, and we will keep talking about them with workers, governments, civil society, and communities across the globe,” the business stated.

After heading the augmented reality glasses team at Meta, Kalinowski moved to OpenAI in November 2024.

In a late February agreement with the Pentagon, OpenAI agreed to use cutting-edge AI technologies in classified settings, and the business asked the government to make these technologies accessible to all AI firms as well, according to a statement.

“We maintain entire control over our safety stack, we deploy via cloud, cleared OpenAI people are in the loop, and we have solid contractual protections. All of this complements the robust current protections afforded by US law,” the business stated.

Following discussions between the Trump administration and Anthropic that fell through, the agreement was struck. The Pentagon’s designation of Anthropic as a supply chain danger will be disputed, according to the corporation.

Additionally, OpenAI publicly disputed the decision to blacklist Anthropic.

Additionally, Anthropic announced that the US government has officially recognized the business as a “Supply Chain Risk (SCR)” and that the CEO has also apologized for disparaging President Donald Trump.

The CEO made it clear that the term would only refer to clients using Anthropic’s Claude models under Department of War contracts, not to all uses of Claude by those clients.