Governo dos EUA diz a empresas de IA que modelos de peso aberto da China serão dispensados de testes

Trump Administration Won’t Safety-Test Open-Weight AI Models as OpenAI and Anthropic Raise Cybersecurity

White House excludes open-weight models from federal AI safety testing amid OpenAI/Anthropic agent breakout incidents. Regulatory move prioritizes open-source competition but widens cybersecurity exposure; China's GLM-5.2 narrows gap with GPT-5.5/Claude 4.7.