中文翻译
摘要: 因美国监管机构出于国家安全担忧,Anthropic在发布AI模型“Claude Fable 5”仅数日后,被迫紧急暂停其全球访问权限。此举主要防范AI“越狱”技术绕过安全限制引发网络滥用。Anthropic已配合政府指令,称相关漏洞有限且业界已知。此次叫停创下AI行业罕见的最快政策回调纪录。
正文
About Us
Trending
Contact Us
Advertise
UAE Edition
Insights
Cryptocurrencies
Magazines
Industry
Geography
Insights
Cryptocurrencies
Magazines
Industry
Geography
US Cracks Down on Anthropic, Halts Claude Fable 5 Within Days of Launch Over Security Fears
Anthropic faces mounting pressure as the US halts global access to Claude Fable 5 over jailbreaking risks, while a legal battle unfolds and regulators flag potential cyber misuse despite limited known vulnerabilities
Written By:
Simran Mishra
Reviewed By:
Manisha Sharma
Published on
13 Jun 2026, 8:04 am
Updated on
13 Jun 2026, 8:04 am
Anthropic has paused access to its most advanced artificial intelligence systems after US security concerns triggered immediate regulatory action. The company confirmed that authorities instructed a halt on foreign access to its newly released Claude Fable 5 and Mythos 5 models.
The decision arrived only days after the public rollout, marking one of the fastest reversals in the AI industry. Anthropic described the move as a compliance step following government directives linked to national security priorities.
Claude Fable 5 had positioned itself as a next-generation system competing with tools like OpenAI’s ChatGPT and
Google’s Gemini
. The company previously highlighted its unmatched capabilities and warned about potential misuse risks.
Officials raised alarms around possible AI jailbreaking techniques that could bypass built-in safeguards. Jailbreaking allows users to override restrictions and access sensitive system functions or hidden capabilities. Anthropic acknowledged awareness of such techniques, although it described the vulnerabilities as limited and already known within the research community.
Concerns Over AI Jailbreaking and Security Risks
The company stated that similar weaknesses appear across several publicly a
来源: Brave/analyticsinsight.net
采集时间: 2026-06-13 20:58:46
AI人工智能AnthropicClaude