[DigitalToday reporter Hyunwoo Choo] OpenAI CEO Sam Altman (샘 알트먼) held an AMA (questions and answers) on X, formerly Twitter, on Saturday local time about an AI contract with the U.S. Department of Defense. OpenAI decided to provide its AI model after Anthropic refused the Defense Department’s demands. Altman said the deal was reached in haste but he expected a positive outcome.

According to a Business Insider report, Altman explained why OpenAI signed with the Defense Department by saying, “It seems Anthropic wanted more operational control, and we reached an agreement quickly as we became familiar with the contract language.”

Altman also said OpenAI had set three “red lines” on the use of its AI models, but added they could change as the technology advances. “We can decide whether ChatGPT answers controversial questions, but major ethical issues such as responding to nuclear bombs should be decided by the government,” he said.

Altman said cooperation with the Defense Department had been discussed informally for several months and recently shifted to classified negotiations. “The current conflict between the AI industry and the government is dangerous for Anthropic, healthy competition and the United States,” he said. He added that OpenAI negotiated so that all AI labs would be offered similar terms.

On how AI could be used, he said it would play a big role in defending against cyber attacks and in biosecurity. He said it would be useful in responding to cyber attacks that paralyse a national power grid and in responding to new pandemics. OpenAI also said it could terminate the contract if the government violates the terms, but added that such a situation was unlikely.

Earlier, the U.S. Defense Department also demanded that Anthropic use AI for large-scale surveillance and the development of autonomous weapons. Anthropic refused, leading to the cancellation of a $200 million defense contract and a ban on cooperation with federal agencies. Anthropic signalled legal action, but criticism has also emerged that the AI industry has fallen into a trap of its own making, IT media outlet TechCrunch reported.

Max Tegmark (맥스 테그마크), a Swedish-American physicist and an MIT professor, points to the outcome as a boomerang effect from AI companies choosing to avoid regulation. Anthropic, OpenAI and Google DeepMind, among others, have emphasised safety while opposing strong regulation.

The analysis is that as the AI industry has grown in a largely unregulated state, companies have ended up unable to refuse government demands. “As AI is seen as a threat to national security, AI companies are now facing risks of their own making,” Tegmark said.

The AI industry is evolving rapidly, but regulation is not keeping pace. AI has advanced to the point of winning elite mathematics competitions, but companies are still calling for self-regulation. Yet as AI becomes directly linked to national security, government intervention has become inevitable. “AI companies argued for self-regulation and blocked government intervention, but now that choice is instead holding them back,” Tegmark said.

Attention is now on whether AI companies will reset relations with government in the wake of the Anthropic case, or accept defense contracts. Google and xAI are staying silent on the Anthropic case.