Steve Bannon says Anthropic 'had it right' in rejecting deal with the Pentagon

During the Semafor World Economy Summit held in Washington, D.C., former White House chief strategist Steve Bannon expressed his support for Anthropic's choice to reject a deal with the Pentagon. He stated, "I think Anthropic had it right," emphasizing the dangers of allowing the military to utilize their advanced AI model, Claude, without sufficient oversight. Bannon, who has been vocal about the risks associated with superintelligent AI, called for increased transparency regarding the application of AI in weapons development. He remarked, "The central thing is what is happening in the weapons lab with AI; we have no earthly idea." The disagreement between Anthropic and the Pentagon escalated earlier this year as negotiations unfolded concerning the military's access to Claude. Defense Secretary Pete Hegseth urged Anthropic to comply with the Pentagon's terms, warning that failure to do so could jeopardize their military contract. In response, Anthropic's CEO, Dario Amodei, articulated the company's ethical reservations, particularly regarding mass domestic surveillance and the deployment of fully autonomous weapons, stating they could not, in good conscience, acquiesce to the Pentagon's demands. Following Anthropic's refusal, the Pentagon reacted by categorizing the company as a supply chain risk, effectively banning federal agencies from utilizing their technology. In March, Anthropic initiated legal action against Hegseth and several federal entities in response to the blacklist. Meanwhile, the Pentagon swiftly forged an agreement with OpenAI, led by Sam Altman. Despite the tumultuous legal and corporate environment, Anthropic managed to emerge favorably in the public perception, with Claude temporarily surpassing ChatGPT in the App Store rankings. The company has recently attracted attention again with its announcement of a new model, Mythos, which has been placed on hold due to cybersecurity issues. In their system card, Anthropic indicated that the model's enhanced capabilities led them to restrict its availability, opting instead to use it collaboratively within a limited defensive cybersecurity framework.