澳大利亚主权大赦国际在就外国技术影响进行辩论时,在当地推出适合本国国情的具有国家数据主权的AI模型。 Sovereign Australia AI launches locally tailored AI models with national data sovereignty amid debate over foreign tech influence.

澳大利亚主权大赦国际是澳大利亚的一个新开办单位,正在建立当地定制的AI模型——Australis(700亿参数模型)和Ginan(较小型的开放源码版本),使用256台Nvidia Blackwell B200 GPUs,这是澳大利亚最大的自主AI硬件部署。 Sovereign Australia AI, a new Australian startup, is building locally tailored AI models—Australis, a 700-billion-parameter model, and Ginan, a smaller open-source version—using 256 Nvidia Blackwell B200 GPUs, the country’s largest sovereign AI hardware deployment. 该公司旨在反映澳大利亚的文化、语言和法律框架,同时在对美国《云法案》的关切中确保数据主权。 The company aims to reflect Australian culture, slang, and legal frameworks while ensuring data sovereignty amid concerns over the US Cloud Act. 它计划支持政府的AI倡议,如Gov AI Chat, 并已向版权持有者认捐1 000万美元。 It plans to support government AI initiatives like Gov AI Chat and has pledged $10 million to copyright holders. 虽然OpenAI在澳大利亚扩展了本地化产品,但主权澳大利亚AI认为外国模型可能与国家价值观不一致. While OpenAI expands in Australia with localized offerings, Sovereign Australia AI argues that foreign models may not align with national values. 批评者,包括阿特拉斯的Mike Cannon-Brookes和技术专家Craig Dargusch警告说,澳大利亚缺乏足够的历史文本数据,应该把重点放在实际的AI应用上,而不是从头建立基础模型。 Critics, including Atlassian’s Mike Cannon-Brookes and tech expert Craig Dargusch, caution that Australia lacks sufficient historical text data and should focus on practical AI applications rather than building foundational models from scratch.