美国防长接近切断五角大楼与Anthropic的商业联系

Core Viewpoint - The U.S. Department of Defense is poised to sever its business relationship with Anthropic, categorizing the AI company as a "supply chain risk," which will require any business wishing to work with the military to cut ties with Anthropic [2][10]. Group 1: Current Situation - The Pentagon's chief spokesperson stated that the partnership with Anthropic is under review, emphasizing the need for partners to support U.S. military operations [2][10]. - Anthropic's AI model, Claude, is currently the only AI model approved for use in military classified systems and has been integrated into military operations, including a raid against Maduro earlier this year [2][10]. Group 2: Negotiation Challenges - Anthropic and the Pentagon have been engaged in intense negotiations over the terms of military use of Claude, with Anthropic willing to relax some restrictions but aiming to prevent its tools from being used for mass surveillance or autonomous weapon development [3][11]. - The Pentagon finds these restrictions too stringent and insists on the ability to use the tools for "all lawful purposes," indicating a significant gap in expectations between the two parties [3][11]. Group 3: Legal and Ethical Concerns - Current laws regarding mass surveillance do not adequately address the implications of AI, raising concerns that AI could enhance the military's ability to monitor civilians [12][13]. - An Anthropic official noted that while the Pentagon has the legal right to collect open-source intelligence, the scale of monitoring could dramatically increase with AI capabilities [13]. Group 4: Potential Impact - If Anthropic is designated as a supply chain risk, numerous companies working with the Pentagon will need to demonstrate that they do not utilize Claude, which could affect a significant number of businesses, as Anthropic claims that eight out of the top ten U.S. companies are using Claude [4][14]. - The Pentagon's threat to cancel contracts with Anthropic could involve up to $200 million, a small fraction of Anthropic's $14 billion annual revenue [5][14]. Group 5: Future Negotiations - The Pentagon's firm stance against Anthropic sets a precedent for negotiations with other AI companies like OpenAI, Google, and xAI, which have agreed to lift protective restrictions for military use in non-classified systems [7][16]. - There is confidence that these companies will accept the "all lawful purposes" standard, although many issues remain unresolved [7][16].