The smoke of the global AI battlefield is spreading from Silicon Valley laboratories to the classified networks of the Pentagon.

According to the latest reports, OpenAI has signed a landmark new agreement to sell the rights to its top AI models through AWS (Amazon Web Services) to the U.S. Department of Defense and government agencies. This means that the GPT series models will officially enter the U.S. defense system, used for core tasks including both classified and non-classified information.

Behind this "fast replacement," another AI giant Anthropic has quietly withdrawn. It is reported that the company, known for its "AI safety," had its cooperation with the government completely broken in February this year. The reason was that Anthropic refused to allow its Claude model to be used without restrictions in military applications, especially in domestic surveillance and autonomous weapons. Subsequently, the Pentagon unusually classified this domestic company as a "supply chain risk."

With Anthropic stepping out, the power vacuum was quickly filled:

Model Revolution: Multiple agencies, including the State Department, have abandoned Claude and turned to their competitors. For example, the underlying model of the internal robot StateChat has been quickly upgraded to GPT-4.1.

Amazon's "High-Stakes Bet": In OpenAI's recent funding round, Amazon pledged a $5 billion investment, surpassing NVIDIA and SoftBank. The two companies also reached an eight-year strategic agreement, with AWS becoming the exclusive third-party distributor for its enterprise platform Frontier.

Currently, Anthropic has filed a lawsuit against the Pentagon, accusing it of illegally labeling it as a "supply chain risk" and violating freedom of speech. However, the business landscape has already been reshaped. By the end of last month, the annualized revenue of OpenAI's enterprise business had reached $10 billion, and it is now working hard to catch up with Anthropic, which currently leads in the enterprise market.