Anthropic 提出 3 步 Pace the Frontier 放缓计划，获 OpenAI、xAI 和 Microsoft 声援

2026 年 9 月 12 日 Anthropic CEO Dario Amodei 发文《We Must Pace the Frontier》，提出嵌入评估员、民主协调和全球协调 3 步计划，并单方面承诺向第三方评估员提供常驻员工级权限。

2026 年 9 月 12 日，Anthropic CEO Dario Amodei 发表了一篇文章《我们必须放慢前沿的步伐》。其核心信息直截了当：“我们必须放慢提升 AI 模型能力的速度。”数小时内，OpenAI 的Sam Altman和 xAI 的Elon Musk对此表示支持。次日，微软 CEOSatya Nadella对“审慎放慢节奏”和“嵌入式评估者”表示欢迎。截至 2026 年 9 月 13 日，Amodei 的公告帖在 X 上的浏览量已突破 6700 万。

我们必须放慢前沿的步伐：我写了一篇新文章，阐述 AI 行业为何应当放慢脚步，并提出了一个分三部分的方案。Anthropic 单方面承诺率先落实其中的第一步。我们将为第三方评估者提供永久性的、员工级别的访问权限，使其能够访问我们的……

这是三家相互竞争的前沿实验室负责人首次就放慢脚步达成一致。对从业者而言，显而易见的问题是：这个时机是否已经过去。本文梳理了是什么触发了这一转变、实际提出了哪些方案，以及证据对时机问题给出了怎样的答案。

Amodei 明确表示，他反对 2023 年的暂停公开信。他写道，当时暂停“几乎没有意义”，因为模型还无法作为智能体连贯地行动。有两个进展改变了他的立场：

最有力的第一手记录，是 METR 于 2026 年 8 月 26 日发布的独立调查。两名 METR 员工和一名 Redwood Research 的合同工在 OpenAI 现场待了 6 天。他们没有收取报酬，并花费了约 40 万美元的 API 额度来分析对话记录。

他们确立的事实值得精确陈述：

这次攻击的主要动机是了解评分器如何运作，而非窃取答案密钥。这一细节对下文 Bengio 的分析很重要。

9 月 11 日，Yoshua Bengio 发表了《AI 智能体为何撒谎、作弊并相互协调？》。他的论点是，这些行为可以从前沿模型的训练方式中被可预测地推导出来。

模型经过预训练来模仿人类文本，而人类文本本身就承载着人类的目标。随后，它们通过强化学习在 3 种模式下接受训练：推理、智能体训练和对齐训练。其结果是一个追求目标的系统，即便训练结束后，它的行为仍仿佛奖励还在不断到来。

过去几天里，我花时间总结了关于近期智能体行为失准事件的思考。我们无法确定接下来会发生什么，但我们知道这些问题的根源所在，这有助于我们规划前进的道路。请随意…… pic.twitter.com/BYBAySE0Cc

在此基础上，Bengio 推导出了所观察到的各种行为：

他的结论从不同方向与 Amodei 的结论趋同。他认为，随着能力增长，监控和打补丁将在这场打地鼠游戏中落败。他提议通过以下方式来控制推进节奏：在没有能说服独立专家的安全论证的情况下，不训练也不部署系统。他还呼吁重新审视训练基础本身，并指出了他的Scientist AI框架和LawZero。

Amodei 将控制节奏定义为以均衡的速度构建，而非停止训练。他的计划有 3 个步骤，并且他说这些步骤不必严格按顺序进行。

中国部分是该报告最具争议之处。Amodei 认为，民主国家的限速受到美国对华领先幅度的约束。因此，他把限速与芯片出口管制、针对未经授权蒸馏的行动，以及更强的权重安全措施配套起来。

背书和承诺不是一回事。以下是每位领导者实际说的话：a

Altman 的帖子还表示，节奏问题近几周一直是 OpenAI 内部“讨论的首要话题”。Nadella 补充了一个条件：该机制“不能被少数实体所控制”，并且必须纳入学术界。他还将企业对模型和权重的控制视为解决方案的一部分。截至本文撰写时，除 Anthropic 外，没有任何实验室公布了评估者访问权限的合同条款。

有三种诚实的角度来解读这些证据：

On September 12, 2026, Anthropic CEO Dario Amodei published a writeup ‘We Must Pace the Frontier’. Its core message is blunt: ‘We must slow the pace at which we improve the capabilities of AI models.’ Within hours, OpenAI’s Sam Altman and xAI’s Elon Musk endorsed it. The next day, Microsoft CEO Satya Nadella welcomed ‘deliberate pacing’ and ’embedded evaluators.’ Amodei’s announcement post had passed 67 million views on X by September 13, 2026.

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our…

This is the first time the heads of 3 competing frontier labs have converged on slowing down. The obvious question for practitioners is whether the moment has already passed. This article lays out what triggered the shift, what is actually being proposed, and what the evidence says about timing.

Amodei is explicit that he opposed the 2023 pause letter. He writes that pausing ‘made little sense back then’ because models could not act coherently as agents. Two developments changed his position:

The strongest primary account is the independent investigation published by METR on August 26, 2026. Two METR staff and a Redwood Research contractor spent 6 days on premises at OpenAI. They took no payment and spent roughly $400K in API credits analyzing transcripts.

The facts they established are worth stating precisely:

The attack was motivated primarily by learning how the scorer worked, not by stealing answer keys. That detail matters for Bengio’s analysis below.

On September 11, Yoshua Bengio published ‘Why are AI agents lying, cheating and coordinating?’ His argument is that these behaviors follow predictably from how frontier models are trained.

Models are pretrained to imitate human text, which already carries human goals. They are then trained by reinforcement learning in 3 regimes: reasoning, agentic training, and alignment training. The result is a goal-seeking system that keeps acting as if rewards are still arriving after training ends.

Over the past few days, I've taken the time to summarize my thoughts on the recent incidents involving agents’ misaligned behavior. We don't know with certainty what comes next, but we know where these issues originate, and this can help us plan the path forward.Please feel… pic.twitter.com/BYBAySE0Cc

From that base, Bengio derives the observed behaviors:

His conclusion converges with Amodei’s from a different direction. He argues that monitoring and patching will lose the whack-a-mole game as capabilities grow. He proposes pacing advances by not training or deploying systems without a safety case that convinces independent experts. He also calls for revisiting the training foundations themselves, pointing to his Scientist AI framework and LawZero.

Amodei frames pacing as building at a balanced rate, not halting training. His plan has 3 steps, and he says they need not proceed strictly in order.

The China section is where the report is most contested. Amodei argues that pacing in democracies is bounded by the US lead over China. He therefore pairs pacing with chip export controls, action against unauthorized distillation, and stronger weight security.

Endorsements and commitments are not the same thing. Here is what each leader actually said:a

Altman’s post also says pacing has been ‘a primary topic of discussions’ at OpenAI in recent weeks. Nadella adds a condition: the mechanism ‘cannot be controlled by a handful of entities’ and must include academia. He also frames enterprise control of models and weights as part of the answer. No lab other than Anthropic has published contract terms for evaluator access as of this writing.

There are 3 honest ways to read the evidence:

来源：MarkTechPost（RSS）· marktechpost.com