Altman defends the Pentagon agreement at the all-hands meeting, saying the rebound "really hurts"; OpenAI is also seeking NATO contracts

According to 1M AI News monitoring, OpenAI CEO Sam Altman defended the Pentagon’s confidential network AI deployment agreement to employees at a company-wide meeting on Tuesday. The agreement was announced last Friday, just hours after Defense Secretary Pete Hegseth labeled a competitor, Anthropic, as a supply chain risk—an action rarely used against domestic U.S. companies. The announcement was met with widespread criticism from internal staff and Silicon Valley AI researchers, who argued that OpenAI’s acceptance of terms allowing AI for “all lawful purposes” was equivalent to conceding to the Pentagon. In response to the backlash, OpenAI amended the agreement to explicitly prohibit use for domestic surveillance.

According to remarks seen by The Wall Street Journal, Altman stated he does not regret signing the agreement but admitted the announcement was too rushed, appearing “opportunistic” and “disunited with the industry.” He said, “We’ve worked so hard to do the right thing, only to be completely crushed on a personal level—I know you’re going through this too, so I’m very sorry for what you’re enduring—it’s really painful.”

Altman also said at the meeting that OpenAI is seeking to deploy AI across all NATO classified networks, but a company spokesperson later clarified that Altman misspoke, and the contract would only target NATO’s “non-classified networks.” Altman expressed that the government is willing to let OpenAI influence how its technology is deployed, and he hopes the company “has a seat at the decision-making table,” emphasizing that “a strong U.S. military has been a tremendous benefit to all of humanity over the past 250 years.”

Additionally, The Wall Street Journal reported that last weekend’s military strikes against Iran involved Anthropic’s Claude model in planning; Deputy Secretary of Defense for Research and Engineering Emil Michael stated at a16z event on Tuesday that AI models’ constitutional framework “must not override the U.S. command and control system.” During the backlash, Anthropic’s Claude app surpassed ChatGPT to become the most popular app on the Apple App Store for the first time. (Wall Street Journal)