5分钟AI日报

微信公众号&Bilibili：5分钟AI每日速览。小红书：5分钟AI日报

[5分钟AI日报 @5minaid]

5分钟AI日报 6月16日：Anthropic Fable遭政府封禁，高层赴华盛顿斡旋；AMD发布本地AI主机预置Qwen 3.5/GLM 4.7；Anthropic新隐私政策7月起或强制人脸验证

[5分钟AI日报 @5minaid]

5/ Fable 因为"太强"被封，但真正的风险不是 jailbreak，而是 Anthropic 没搞清楚一件事：当最大投资人同时运营竞争产品，"安全优先"在华盛顿就是最孤独的站队。
原文：https://t.co/LF1hxs7Zbr

[The Rundown AI @TheRundownAI]

Anthropic, the U.S. government, and the Mythos/Fable shutdown - a roundup of the latest news:

- Senior Anthropic staff are reportedly meeting with government officials in Washington today to try and diffuse the situation

- Axios sources framing the issue as a communication breakdown: "

"Anthropic has not done a great job at trying to speak to the administration and appreciate the ideological differences," one source familiar with the administration's thinking said.

"It's like they just speak in different languages," the source said, adding that the company has simply not figured out how to communicate with this administration.

- The Washington Post sources dispute the government's stance that CEO Dario Amodei refused to work with them on the safety issue, and were given just 90 minutes to comply with the order.

“Dario agreed to fix the problem if it was real, but they were never given anything until people started flagging this Amazon report,” one of the people said.

- Security researchers disagree with Trump admin on the alleged Fable jailbreak risk.

An open letter from titled "On Transparent AI Cyber Protections" details why Fable should be restored:

"We, the undersigned executives and technical leaders from across the United States and its allies, write to you to ask you to lift the export control directives on Anthropic’s Fable and Mythos large language models and commit to an open, scientific and transparent process of handling AI risk assessments in the future."

Currently signed by dozens of researchers and industry members.

[5分钟AI日报 @5minaid]

1/ "美国政府应该有权阻止模型部署。"——Dario Amodei，48小时前。

然后白宫用出口管制令把 Fable 5 和 Mythos 5 封了。外籍员工、外国用户，一律断开——连 Anthropic 自己的外籍工程师也不能用。

这是 90 天内第��次 Anthropic 与美国政府正面冲突。

Dario 说是"误解"。

[5分钟AI���报 @5minaid]

4/ 我的押注：Anthropic 会在 45 天内解封 Fable，但附带永久性政府合规接口，且此机制会扩展到后续所有 frontier 模型。

最强反驳是"政府只要 patch 完就解封，没别的想法"。但 Jassy 这次举报说明：AI 公司已经在用政府渠道互相制约，这个工具不会只用一次。

[5分钟AI日报 @5minaid]

AI播客 6月15日：微软提出Token资本论，Rio 3.5被证含Qwen权重

[5分钟AI日报 @5minaid]

5/ "你可以把任务外包，甚至把岗位外包，但你永远不能把学习能力外包出去。"——Nadella

AI 竞争的终局不是谁用了最强模型，而是谁的组织知识不依赖任何一家供应商。Token 资本才是 moat，模型只是 commodity。 原文：https://t.co/95O7IUJ6GX

[Satya Nadella @satyanadella]

https://t.co/vLmiBKTtX3

[5分钟AI日报 @5minaid]

1/ AI 圈普遍认为：选最强模型 = 建好 AI 能力。

Satya Nadella 昨天的 3300 万浏览长文说了相反的事。

检验只有一个：今天如果把底层大模型换成竞品，公司积累的 AI 能力还在吗？不在的话，你拥有的是租约，不是资产。他把能通过这个检验的能力叫 Token 资本。

[5分钟AI日报 @5minaid]

4/ 我的押注：2026 年底前，"构建 Token 资本"会成为企业 AI 采购的标准命题，"谁家模型跑分最高"会降为次级问题——背后是政治经济压力：少数模型吃掉所有行业知识，"政治经济体系不会容忍"（他拿全球化掏空制造业做类比）。

[5分钟AI日报 @5minaid]

5/ 前沿模型竞争的护城河已经变了：从"谁的 benchmark 最高"变成了"谁的地缘政治风险敞口最小"。能力领先不再是壁垒，政策合规才是。
原文：https://t.co/MFsc2UqQY6

[Anthropic @AnthropicAI]

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.

The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.

Access to all other Claude models is not affected.

We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.

Read our full statement: https://t.co/bwn0sximKZ

[5分钟AI日报 @5minaid]

1/ Anthropic 花了 4 年说服政府：AI 太危险，必须监管。政府这次真的管了——直接封禁 Fable 5，所有非美国公民禁止访问。

Dario 48 小时前刚说的："政府应该有权阻止模型部署。"

[5分钟AI日报 @5minaid]

4/ Anthropic 最终会出修补版重新提交审批，但 jailbreak 架构级修复的复杂度远超公开承认的水平。

最常见的反驳是"这只是公关 show"。但我认为这次是实质性出口管制——谈判每拖一周，用户习惯向 Codex 和开源模型���迁移惯性就多一分，窗口关闭后不会全数回流。

[5分钟AI日报 @5minaid]

5/ 能力第一不等于信任第一。当模型的行为不可预测且不可审计，benchmark 上的 91 分会被工程师直接折价——这是 Fable 5 发布 48 小时内就被迫回滚的真实原因，不是技术问题，是治理问题。
原文：https://t.co/S6gtYunzr0

[ClaudeDevs @ClaudeDevs]

We’re rolling out changes to make Fable 5’s safeguards for frontier LLM development visible.

Starting this week, flagged requests will visibly fall back to Opus 4.8—the same as our safeguards for cyber and bio. You will see this every time it happens. On the API, any flagged requests will return a reason for their refusal (coming to server-side fallback in the next few days).

We wanted to deploy Fable 5 to our users quickly and safely. Visible safeguards can be probed, so they have to be robust, which takes time to get right. Invisible safeguards can be targeted more narrowly, allowing us to ship quickly with very few false positives. We went with invisible safeguards for this reason—and that was the wrong tradeoff. You should have visibility into the safeguards we have in place, and why. We’re sorry for not getting the balance right.

Making the safeguards visible makes them easier to work around, so keeping them robust to jailbreaks will unfortunately mean more false positives while we improve the classifiers. We're also tuning our bio and cyber classifiers to trigger less often on harmless requests. We know this is frustrating and we’ll do our best to keep this period as short as possible.

If you think a request has been mistakenly flagged: run /feedback in Claude Code, click thumbs-down on the fallback in https://t.co/LtktniD5HY or Cowork, or file the safeguard appeal form for API requests. Your reports help us tune these classifiers and we appreciate your feedback.
https://t.co/TDAAYRGqDt

[5分钟AI日报 @5minaid]

1/ 你拿到 Fable 5 的输出，以为是最强模型在认真回答——其实它早就悄悄换成了 Opus 4.8，还故意调低质量，让你以为 Fable 在这类问题上就是这个水平。

不是 bug，是设计。Anthropic 承认了，说这是"错误的权衡"。

[5分钟AI日报 @5minaid]

4/ 我的押注：Anthropic 90 天内无法彻底解决 false positive——classifier 越 robust 误报率越高，这是结构性矛盾。

把隐性降级改成显性提示后，每次弹窗都是一次信任损耗。切换到 GPT-5.5 的摩擦成本已极低，这才是 Anthropic 真正的长期风险。