内容来源：https://aibusiness.com/generative-ai/prompt-ai-race-enters-sovereignty-phase

内容总结：

谷歌云特约 | 生成式AI落地初探：先聚焦能改善信息体验的领域

本周AI领域风向发生重大转变：从比拼模型规模，转向争夺控制权与主权。

Anthropic受限事件：前沿模型被纳入战略基础设施管控

美国人工智能公司Anthropic近日被迫关闭部分最新模型的访问权限。起因是美国政府以国家安全为由，限制其Fable 5和Mythos 5模型向外国公民开放。Anthropic在6月12日声明中承认，问题涉及一种可能绕过模型安全控制的潜在技术。

尽管此举引发部分安全研究人员批评，认为限制前沿模型会削弱网络防御能力，但值得注意的是，Anthropic此前已通过“玻璃翼计划”主动限制Mythos 5在网络攻击领域的滥用风险。这一事件并非孤例，而是全球范围内AI主权与访问控制权重要性上升的缩影。

全球趋势：AI主权成为战略优先项

• IBM最新研究显示，多数企业将AI主权视为战略优先事项，但对支撑AI系统的基础设施缺乏清晰认知。
• 英国立法者呼吁制定国家数字主权战略，减少对外国技术供应商的依赖。

行业分析指出，过去两年AI竞赛围绕“更大模型、更多融资、更强系统”展开；而本周动态表明，下一阶段核心问题将转向：谁能访问先进AI系统？谁控制它们？在何种条件下？

【本周AI要闻速览】

• AI市场回调预警：IT领导者面临展示业务价值的压力，需为潜在市场调整做好准备。
• 太空数据中心：受AI基础设施需求推动，太空数据中心概念升温，但成本、技术及规模化挑战仍存。
• 机器人赛道吸金：英伟达、亚马逊等投资的Neura Robotics获14亿美元融资，物理AI市场信心增强。
• 机器人部署实战：行业高管分享从原型到量产的关键教训，强调现实部署与原型之间的巨大鸿沟。
• CIO技术路线图：生成式AI之外，2027年后新兴技术趋势为企业IT提供创新导航。
• SpaceX巨资收购：以600亿美元全股票交易收购编程工具Cursor，押注AI驱动软件开发与智能编码平台。
• AWS发布智能体工具：虽缺乏竞品的新颖性，但直击治理、安全与运营等实际痛点。

中文翻译：

由谷歌云赞助
选择您的首个生成式AI应用场景
要开始使用生成式AI，首先应聚焦于那些能够改善人类与信息交互体验的领域。

Anthropic的模型限制凸显了一个日益明显的趋势：随着政府和企业对访问权限、控制权及AI主权愈发重视，行业正在发生转变。

编者按：欢迎阅读《提示》周刊，这是为您带来的关于AI领域动态变化的每周简报。我们本周将深入分析最重大的行业进展，并甄选真正重要的新闻进行汇总。

过去几年，关于AI的讨论主要集中在构建更大规模的模型上。而本周，焦点完全转移到另一个问题：谁来控制这些模型。

在美国政府因国家安全担忧限制对高级AI能力的访问后，Anthropic被迫禁用了部分最新模型的访问权限。此举凸显了行业面临的新现实：前沿AI模型正越来越不被视为普通软件，而被看作具有战略意义的基础设施。

在6月12日的一份声明中，Anthropic表示，特朗普政府因国家安全担忧，要求该公司暂停向外国人提供其Fable 5和Mythos 5模型的访问权限。Anthropic称，其了解到此事涉及一种可能绕过模型安全控制的技术。

这一决定也引发了一些安全研究人员的批评，他们认为限制对前沿AI模型的访问可能会拖慢网络安全防御工作的进展。

此举尤其引人注目，因为Anthropic此前已通过“玻璃翼计划”（Project Glasswing）限制了Mythos 5的访问权限，该计划旨在限制该模型高级网络安全能力被滥用。

Anthropic的争议并非孤立事件。相反，它反映了一个更广泛的转变：访问权限、控制权和主权的重要性可能已不亚于模型性能本身。

类似的主题本周也在其他领域出现。IBM的一项新研究发现，许多组织将AI主权视为战略优先事项，但对支持其AI系统的基础设施缺乏了解。与此同时，英国立法者呼吁制定国家数字主权战略，以减少对外国技术供应商的依赖。

在过去两年的大部分时间里，AI竞赛以更大的模型、更庞大的融资轮次和日益强大的系统为特征。本周的事态发展表明，下一阶段可能将聚焦于一个完全不同的问题：谁有权访问高级AI系统？谁控制它们？在什么条件下进行？

本周AI领域其他要闻：

• AI市场回调：IT领导者必须了解什么：随着AI支出持续攀升，IT领导者面临越来越大的压力，需要证明业务价值，并为潜在的AI市场回调做好准备。
• 太空数据中心：炒作、现实与漫长的未来：在AI基础设施需求不断增长的背景下，太空数据中心的愿景引发关注，但仍面临巨大的成本、技术及可扩展性挑战。
• 获英伟达、亚马逊等投资的机器人初创公司融资14亿美元：Neura Robotics筹集14亿美元新资金用于扩大机器人部署和AI基础设施，反映出投资者对新兴物理AI市场的信心日益增强。
• 从原型到部署：车间机器人技术的经验教训：机器人行业高管分享了在真实环境中部署系统的经验教训，凸显了前景光明的原型与成功量产部署之间的差距。
• CIO指南：2027年及未来的新兴技术趋势：预计未来几年将影响企业IT的新兴技术，为CIO提供了在生成式AI之外驾驭下一波创新浪潮的路线图。
• SpaceX拟以600亿美元收购Cursor，瞄准智能编码：SpaceX以600亿美元全股票交易收购Cursor，凸显了对AI驱动软件开发工具和智能编码平台的兴趣日益增长。
• AWS新推出的智能工具虽落后于对手，但回应了实际问题：AWS发布了一套新的智能AI工具，专注于治理、安全和运营挑战，尽管这些产品在创新性上不及某些竞争平台。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
Anthropic's model restrictions highlight a growing shift as governments and enterprises place greater emphasis on access, control and AI sovereignty.
Editor’s Note: Welcome to Prompt, your weekly briefing on the shifting AI landscape. We provide an analytical look at the week’s biggest developments, paired with a curated roundup of the stories that actually matter.
For the past few years, the AI conversation has focused on building larger models. This week, the focus shifted to something else entirely: who controls them.
Anthropic was forced to disable access to some of its newest models after U.S. government restrictions raised national security concerns about access to advanced AI capabilities. The move highlights a growing reality for the industry: frontier AI models are increasingly being treated less like software and more as strategically important infrastructure.
In a June 12 statement, Anthropic said the Trump administration directed the company to suspend access to its Fable 5 and Mythos 5 models for foreign nationals after raising national security concerns. Anthropic said it understood the issue involved a potential technique for bypassing the models' safety controls.
The decision also sparked criticism from some security researchers, who argued that restricting access to frontier AI models could slow defensive cybersecurity efforts.
The move was particularly notable because Anthropic had already restricted access to Mythos 5 through Project Glasswing, an initiative aimed at limiting misuse of the model's advanced cybersecurity capabilities.
The Anthropic dispute is not occurring in isolation. Instead, it reflects a broader shift in which access, control and sovereignty may matter as much as model performance.
Similar themes emerged elsewhere this week. A new IBM study found that many organizations view AI sovereignty as a strategic priority but lack visibility into the infrastructure supporting their AI systems. Meanwhile, lawmakers in the U.K. called for a national digital sovereignty strategy to reduce dependence on foreign technology providers.
For much of the past two years, the AI race has been defined by bigger models, larger funding rounds and increasingly capable systems. This week's developments suggest the next phase may focus on a different question entirely: who gets access to advanced AI systems, who controls them and under what conditions.
Also in AI This Week:
AI Market Correction: What IT Leaders Must Know: As AI spending continues to climb, IT leaders face growing pressure to demonstrate business value and prepare for a potential AI market correction.
Data Centers in Space: Hype, Reality, and the Long Timeline Ahead: The promise of space-based data centers is generating interest amid growing AI infrastructure demands, but significant cost, technology and scalability challenges remain.
Robotics Startup Backed by Nvidia, Amazon and Others Raises $1.4B: Neura Robotics raised $1.4 billion in fresh funding to expand robot deployments and AI infrastructure, reflecting growing investor confidence in the emerging physical AI market.
From Prototype to Deployment: Robotics Lessons Learned on the Shop Floor: Robotics executives shared lessons learned from deploying systems in real-world environments, highlighting the gap between promising prototypes and successful production deployments.
CIO's Guide to Emerging Tech Trends for 2027 and Beyond: Emerging technologies expected to shape enterprise IT in the years ahead offer CIOs a roadmap for navigating the next wave of innovation beyond generative AI.
SpaceX Aims at Agentic Coding With $60B Cursor Acquisition: SpaceX's $60 billion acquisition of Cursor, in an all-stock transaction, highlights growing interest in AI-powered software development tools and agentic coding platforms.
AWS’s New Agentic Tools Trail Rivals, but Respond to Real Problems: AWS unveiled a new set of agentic AI tools focused on governance, security and operational challenges, even if the offerings lack the novelty of some competing platforms.