2026年6月30日 美国东部时间晚9:27 / 哥伦比亚广播公司（CBS）新闻
人工智能巨头Anthropic周二表示，联邦政府已解除针对其高性能Claude Fable 5和Mythos 5模型的一系列限制，终结了特朗普政府与该AI公司之间持续一周多的争端。
Anthropic在一份社交媒体帖文中表示，公司将于周三恢复相关模型的访问权限。
美国商务部长霍华德·卢特尼克在X平台上写道，近几周来，其团队“与Anthropic密切合作，对Fable 5进行分析并获批，以确保全美政府层面的步调一致，并强化美国在人工智能领域的领导地位”。
CBS新闻已联系Anthropic和美国商务部，寻求更多细节。
Anthropic于本月早些时候向公众发布了Claude Fable 5，公司称其配备了安全防护措施，可降低该模型被用于网络攻击或其他恶意用途的风险。该公司表示，Mythos 5——一款安全防护措施更少的版本——最初将仅提供给少数大型企业进行测试。
就在Fable 5发布几天后，Anthropic将其下架。该公司称，联邦政府出台的出口管制要求其禁止外国国民访问该模型——这项范围广泛的指令实际上迫使Anthropic停用该模型。
Anthropic在当时的一份声明中表示，政府的担忧似乎集中在一种潜在的“越狱”，即诱骗人工智能模型绕过其安全防护的技术。该公司辩称其安全防护措施有效，并称通过所谓越狱手段发现的漏洞“相对简单”，其他AI模型也可复制此类漏洞。
“[我们]不认为发现了一处有限的潜在越狱漏洞，就应该召回已向数亿用户部署的商用模型，”Anthropic当时表示，“如果全行业都采用这一标准，我们认为这实际上将导致所有前沿AI模型提供商停止推出新模型。”
上周，CNBC和《华尔街日报》报道称，美国商务部已开始允许部分企业和政府机构重新使用Mythos 5，这是对严格出口管制的部分放宽。
Fable 5和Mythos 5被中断访问超过两周之际，特朗普政府正在应对人工智能带来的潜在风险。人们愈发担忧，功能日益强大的AI模型可能会成为寻找软件漏洞的网络犯罪分子的助力，但政府的许多盟友对监管该行业持谨慎态度。
本月早些时候，特朗普总统签署了一项行政令，旨在与联邦政府合作，为最先进的私营AI模型建立一个自愿性的30天审查流程。特朗普此前推迟签署该行政令，告诉记者他不想“阻碍”美国在AI行业的领导地位，也不想给中国留下赶超的机会。
周二，就在Anthropic宣布解除限制前不久，白宫办公厅主任苏西·怀尔斯感谢了那些配合该行政令的企业，包括在“先进模型访问、安全防护测试和安全保障”方面提供合作的企业。
今年早些时候，Anthropic与五角大楼发生冲突，该公司推动制定正式的安全防护措施，以防止军方将其Claude模型——已部署在军方机密系统中——用于全自动武器或对美国民众进行大规模监控。
在双方未能达成协议后，特朗普下令所有联邦机构停止使用Anthropic的AI技术，国防部长皮特·赫格斯瑟将其称为“供应链风险”，试图阻止联邦承包商在军事工作中使用Claude。本届政府将长期支持AI政府监管的Anthropic描述为出于意识形态动机。
Anthropic提起诉讼，一名联邦法官裁定这些限制措施违宪，称其为“奥威尔式”行为，并试图“摧毁”该公司。联邦政府正在上诉。
Anthropic says Trump administration lifted restrictions on some of its most powerful Claude AI models
June 30, 2026 9:27 PM EDT / CBS News
Artificial intelligence giant Anthropic said Tuesday the federal government has lifted a set of restrictions on its powerful Claude Fable 5 and Mythos 5 models, resolving a weekslong dispute between the Trump administration and the AI company.
Anthropic said in a social media post that it will begin restoring access Wednesday.
Commerce Secretary Howard Lutnick wrote on X that in recent weeks, his team has “worked closely with Anthropic to analyze and approve Fable 5 to ensure alignment across the US Government and strengthen America’s leadership in AI.”
CBS News has reached out to Anthropic and the Commerce Department for additional details.
Anthropic released Claude Fable 5 to the public earlier this month, with safeguards the company said would reduce the risk of it being misused for cyberattacks or other nefarious purposes. It said Mythos 5 — a version with fewer guardrails — would initially be made available to a select group of major companies for testing purposes.
Just days after Fable 5 was released, Anthropic took it down. The company said the federal government had issued export controls that required it to block access to the model by foreign nationals — a broad order that Anthropic said effectively required it to disable the model.
In a statement at the time, Anthropic said the government’s concerns appeared to focus on a potential “jailbreak,” or a technique for tricking the AI model and bypassing its guardrails. The company argued its safeguards were effective, and called the vulnerabilities discovered using the apparent jailbreak “relatively simple” and replicable with other AI models.
“[We] disagree that the finding of a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people,” Anthropic said at the time. “If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers.”
Last week, CNBC and The Wall Street Journal reported that the Commerce Department had begun allowing certain companies and government entities to begin using Mythos 5 again, a partial easing of the strict export controls.
The more than two-week-long interruption for Fable 5 and Mythos 5 came as the Trump administration navigates the potential risks posed by artificial intelligence. Fears have mounted that increasingly powerful AI models could be a boon for cybercriminals hunting for software vulnerabilities, but many administration allies are wary of government regulation of the industry.
Earlier this month, President Trump signed an executive order meant to create a voluntary 30-day review process for the most advanced private AI models, in collaboration with the federal government. Mr. Trump had delayed signing the order, telling reporters he didn’t want to “get in the way” of U.S. leadership of the AI industry or risk giving China an edge.
On Tuesday, shortly before Anthropic announced the restrictions were lifted, White House chief of staff Susie Wiles thanked companies that had cooperated with the executive order, including on “advanced model access and guardrail testing and security.”
Earlier this year, Anthropic clashed with the Pentagon over the company’s push for formal guardrails to prevent the military from using its Claude models — which were deployed in the military’s classified systems — to power fully autonomous weapons or mass surveillance of Americans.
After the two sides failed to reach an agreement, Mr. Trump ordered all federal agencies to stop using Anthropic’s AI technology, and Defense Secretary Pete Hegseth labeled it a “supply chain risk,” seeking to stop federal contractors from using Claude for military work. The administration cast Anthropic — long a backer of government oversight of AI — as ideologically motivated.
Anthropic sued, and a federal judge blocked those restrictions, calling them “Orwellian” and an effort to “cripple” the company. The federal government is appealing.
发表回复