"Some have tried to frame distillation as harmful, but I think it is important to protect the principle that you can learn from anything you can observe."
分歧的起点是定性，不是事实
"Some have tried to frame distillation as harmful."
扎克伯格没有否认「模型从模型学习」在发生，他否认的是这件事该被叫作有害。两边描述同一动作，一边用刑事语言，一边用认识论语言。
Anthropic 把安全看作产品属性
"Models built through illicit distillation are unlikely to retain those safeguards."
能力和防护是打包出厂的，蒸馏走能力却剥掉防护，所以危险。这个框架下「负责任」= 守住自己模型的出口。
扎克伯格把安全看作权力结构
"There is no such thing as a singular benevolent superintelligence."
人类不是单一文化，所以不存在能对所有人仁慈的单一模型；安全只能来自谁都别独大。这个框架下「负责任」= 别把能力独占。
最尖锐的一处：互指对方是头号风险
"The most dangerous scenario from this perspective would be leading AI labs training powerful models and keeping them for themselves."
Anthropic 的头号风险是能力流到不该去的地方；扎克伯格的头号风险是能力留在不该独占的地方。他那句「无论用责任和安全怎么合理化」几乎是点名。
但两边都支持芯片出口管制
"Export controls on silicon have been successful for slowing the progress of foreign labs during this critical period."
这不是「开放派 vs 管制派」的对立。真正的争点是管制该卡在哪一层：Anthropic 想延伸到模型访问与蒸馏行为，扎克伯格坚持只卡硬件。
对齐的对象被换掉了
"We view alignment as ensuring that agents share a person's goals and values, not our company's."
Anthropic 的 safeguards 默认用户可能是威胁；扎克伯格的 alignment 默认公司可能是威胁。同一个词，防的对象相反。
But distillation can also be used for illicit purposes: competitors can use it to acquire powerful capabilities from other labs in a fraction of the time, and at a fraction of the cost, that it would take to develop them independently.
但蒸馏也可以用于非法目的：竞争对手能靠它从别的实验室拿走强大能力，所花的时间和成本都只是自主研发的零头。
Some have tried to frame distillation as harmful, but I think it is important to protect the principle that you can learn from anything you can observe.
有人试图把蒸馏描述成有害的，但我认为「你可以从任何你能观察到的东西中学习」这条原则值得守护。
All AI models are derived from human knowledge.
所有 AI 模型都源自人类知识。
Illicitly distilled models lack necessary safeguards, creating significant national security risks.
非法蒸馏出来的模型缺少必要的安全防护，会带来重大的国家安全风险。
Models built through illicit distillation are unlikely to retain those safeguards, meaning that dangerous capabilities can proliferate with many protections stripped out entirely.
通过非法蒸馏造出来的模型不太可能保留这些防护，意味着危险能力会在防护被整块剥掉的状态下扩散。
There is no such thing as a singular benevolent superintelligence.
根本不存在「唯一的仁慈超级智能」这种东西。
The best and most realistic path to building a positive AI future is by delivering superintelligence to everyone.
通向正面 AI 未来的最佳、也最现实的路径，就是把超级智能交付给每一个人。
If distilled models are open-sourced, this risk multiplies as these capabilities spread freely beyond any single government's control.
如果蒸馏出的模型被开源，风险会成倍放大，因为这些能力会自由扩散，超出任何一个政府的控制。
Open source is a positive and important force for empowering people and preventing centralization that is detrimental for both safety and the economy.
开源是一股正面而重要的力量，它赋权于人，并防止那种对安全和经济都有害的中心化。
On cybersecurity, widely deployed open source systems have proven more secure because more people can identify vulnerabilities, harden the systems, and easily upgrade to the latest most secure versions.
在网络安全上，被广泛部署的开源系统已被证明更安全，因为有更多人能发现漏洞、加固系统，并方便地升级到最新最安全的版本。
Distillation attacks therefore reinforce the rationale for export controls: restricted chip access limits both direct model training and the scale of illicit distillation.
所以蒸馏攻击反而强化了出口管制的理由：芯片受限，既限制直接训练模型，也限制非法蒸馏能做到多大规模。
Export controls on silicon have been successful for slowing the progress of foreign labs during this critical period, so it is the right strategic move to continue those.
针对芯片的出口管制在这一关键时期成功拖慢了外国实验室的进展，所以继续维持是正确的战略选择。
Anthropic and other US companies build systems that prevent state and non-state actors from using AI to, for example, develop bioweapons or carry out malicious cyber activities.
Anthropic 和其他美国公司会建立机制，防止国家级与非国家级行为体利用 AI 去做诸如研制生物武器、发动恶意网络活动这类事。
Most labs today view alignment as a defensive measure for enforcing a centralized set of values.
今天大多数实验室把对齐看作一种防御措施，用来强制推行一套中心化的价值。
Instead, we view alignment as ensuring that agents share a person's goals and values, not our company's.
我们的看法不同：对齐应当是确保 agent 认同这个人的目标和价值，而不是我们公司的。
For example, one leading model was aligned to refuse helping draft a letter to prospective parents at a school because it thought standardized testing was unethical.
举个例子，某个领先模型被对齐成拒绝帮忙起草一封给学校准家长的信，理由是它认为标准化考试不道德。
Foreign labs that distill American models can then feed these unprotected capabilities into military, intelligence, and surveillance systems—enabling authoritarian governments to deploy frontier AI for offensive cyber operations, disinformation campaigns, and mass surveillance.
蒸馏了美国模型的外国实验室，可以把这些没有防护的能力送进军事、情报和监控系统——让威权政府把前沿 AI 用在攻击性网络行动、虚假信息宣传和大规模监控上。
While there are risks to releasing capable models, the most dangerous scenario from this perspective would be leading AI labs training powerful models and keeping them for themselves.
发布有能力的模型确实有风险，但从这个视角看，最危险的情形是领先的 AI 实验室训练出强大模型却自己扣着不放。
Regardless of how much a lab rationalizes this activity in terms of responsibility and safety, this is the path of developing a singular superintelligence that cannot be checked by other systems.
无论一家实验室用「责任」和「安全」把这种做法合理化到什么程度，这条路通向的都是一个无法被其他系统制衡的单一超级智能。