Sam Altman Responds to Signing with the US Department of Defense: Hasty Collaboration to De-escalate Situation, AGI Should Possibly be Government-led

BlockBeats News, March 2nd. Yesterday, OpenAI CEO Sam Altman held a public AMA on X, responding to community questions about OpenAI's contract with the US Department of Defense. The original post received over 6.6 million views and more than 7500 replies. Regarding why the deal was rushed, Altman explained that OpenAI had only been in non-confidential discussions with the Department of Defense for months prior. OpenAI had previously turned down contracts in the classified space (which were later undertaken by Anthropic), but after Anthropic was banned, the Department of Defense significantly accelerated classified deployment plans. Altman stated that the rush to sign the contract was an attempt to "de-escalate the situation" and assured that negotiated terms would be open to all other AI labs.

When asked why he did not speak up for Anthropic, Altman classified Anthropic as a "supply chain risk" that is "very bad for the industry, for the country, and for Anthropic." He criticized the Department of Defense's decision and expressed his hope for a retraction. However, he also mentioned that Anthropic seemed more concerned about specific prohibitions in the contract rather than citing existing law in negotiations, suggesting they may want more operational control than OpenAI.

Regarding OpenAI's own red lines, Altman stated: "If asked to do something unconstitutional or illegal, we will withdraw. Please visit me in jail." When discussing foreign surveillance, Altman admitted he "dislikes" the US military's surveillance of foreigners, emphasizing that his most valued AI principle is "democratization," which surveillance may contradict. However, he said, "I don't think it should be up to me to decide."

In his closing remarks, Altman raised a question "hidden behind many questions" that no one had directly asked: What if the US government attempts to nationalize OpenAI or other AI projects? He mentioned that he has "long thought that building AGI might be a government project."