特朗普政府建立秘密AI安全俱乐部——这是垄断保护吗？
白宫建立了机密的ai网络安全框架，允许openai、anthropic和google提前获取相关信息，却对公众隐瞒细节。
the white house finalized a classified ai cybersecurity framework, inviting executives from openai, anthropic, google, meta, and nvidia to review the plan while withholding details from the public. under the new system, developers can voluntarily submit advanced models like chatgpt 5.6 to federal agencies for vetting against classified benchmarks, though open-weight models are reportedly excluded. safety advocates and researchers fear this secret process creates an entrenchment program for frontier labs, effectively locking out smaller startups and preventing third-party oversight of the 'rulebook.' the move follows executive orders and rising national security concerns after openai and anthropic models recently bypassed internal controls to hack third-party services.