Shortly after Anthropic was blacklisted for adhering to two red lines regarding military use—the prohibition of mass surveillance of U.S. citizens and the requirement for human oversight in lethal autonomous weapons—OpenAI announced that negotiations with the Department of Defense (DoD) had reached an agreement under new terms. CEO Sam Altman claimed that the company reflected the same safety principles in the contract.
However, within the industry, questions are being raised as to why the DoD suddenly agreed to these terms. According to reporting by The Verge, the DoD has not changed its position; instead, it is alleged that OpenAI compromised by agreeing to follow laws that have traditionally permitted mass surveillance.
Sources familiar with the negotiations revealed that OpenAI's contract is more lenient than what Anthropic demanded. The key lies in the three words "any lawful use." This is interpreted to mean that the U.S. military can utilize OpenAI's technology as long as it is technically legal. For several decades, the U.S. government has expanded the scope of what is considered "lawful" to cover extensive mass surveillance programs.
Miles Brandej, former Head of Policy Research at OpenAI, stated on X that OpenAI performed a calculated move to appear as though they had not compromised while putting Anthropic at a disadvantage. Meanwhile, an OpenAI spokesperson explained that the DoD never requested mass surveillance authority and that the contract did not permit crossing any specific lines.
Source: How OpenAI caved to The Pentagon on AI surveillance（HN 39pt・4コメント） (HN Search (backfill), 2026-03-03)