内容来源：https://www.geekwire.com/2026/amazon-pushes-back-on-financial-times-report-blaming-ai-coding-tools-for-aws-outages/

内容总结：

亚马逊罕见公开驳斥《金融时报》AI故障报道，称其报道存在事实错误

近日，《金融时报》一篇报道引发科技界广泛关注。该报道援引知情人士称，亚马逊自主研发的AI编程工具Kiro在近几个月内至少导致两次亚马逊云服务（AWS）中断。报道被多家媒体转载，并在知名科技新闻聚合平台Techmeme置顶长达七小时。

上周五下午，亚马逊云科技部门发布题为《关于〈金融时报〉对AWS、Kiro和AI的不实报道的澄清》的博客文章，以异常尖锐的措辞公开反驳相关报道。

亚马逊在回应中承认，去年12月曾在单一区域发生一次有限服务中断，但强调根本原因是“用户在配置访问控制时出现操作失误”，而非AI工具本身缺陷。公司表示：“该问题源于角色配置错误——此类问题在任何开发工具（无论是否AI驱动）或人工操作中均可能发生”，并称未收到任何客户相关问询。

对于《金融时报》所称的“第二次事件影响AWS服务”，亚马逊予以全盘否认。AWS发言人后续向GeekWire补充说明，第二次事件根本未发生在AWS业务范围内，《金融时报》该部分报道“完全错误”。

尽管双方对事件次数存在争议，但去年12月AWS某系统曾中断13小时属实。据《金融时报》报道，四名知情人士透露，当时工程师授权Kiro自主执行变更操作，该AI工具判定最佳解决方案为“删除并重建环境”。多名亚马逊员工称，这是近几个月来AI工具第二次牵涉服务中断。报道援引AWS高级员工说法称，这些中断“规模虽小但完全可预见”，工程师在过程中未进行人工干预。

作为亚马逊最盈利的部门，AWS上季度营收达356亿美元，同比增长24%，运营收入125亿美元。该公司今年计划投入的2000亿美元资本支出中，云业务尤其是AI基础设施是重点投向。

值得关注的是，亚马逊不仅在内部使用此类自主AI工具，也向AWS客户提供相关服务，这使得任何关于AI引发故障的叙事都显得尤为敏感。

亚马逊在回应中重申，12月事件本质是“用户操作失误，而非AI失误”——这一立场虽已出现在《金融时报》原报道中，但亚马逊通过官方博客以更醒目、更尖锐的方式再次强调。公司透露，受影响的服务仅限于39个地理区域中某一区域的AWS成本管理工具Cost Explorer，未波及计算、存储、数据库等核心服务。路透社与The Verge报道指出，受影响区域位于中国大陆。

亚马逊表示已实施新的防护措施，包括对生产访问实施强制同行评审。《纽约时报》记者Mike Isaac在X平台评论称，这是其多年来所见亚马逊“最具攻击性”的公开回应，堪比前白宫新闻秘书、亚马逊公共政策负责人Jay Carney曾为公司强力辩护的风格。

中文翻译：

在科技新闻聚合网站Techmeme榜首盘踞七小时，显然让亚马逊难以忍受。这家科技巨头的云部门亚马逊网络服务（AWS）于周五下午发表了一份异常尖锐的公开驳斥，回应被广泛引用的《金融时报》报道。该报道声称亚马逊自家的人工智能编码工具在最近几个月至少导致了两起AWS服务中断事件。

这则报道被众多媒体转载，并被这家备受关注的科技新闻聚合网站作为案例，用以说明部署自主人工智能工具的风险，以及当出现问题时该由谁（或什么）负责的根本性问题。

在一篇题为"纠正《金融时报》关于AWS、Kiro和AI的报道"的博客文章中，亚马逊承认去年12月在其一个区域的一项服务出现了有限的中断，但将其归因于用户在配置访问控制时的错误，而非AI工具本身的缺陷。

亚马逊表示："问题源于一个配置错误的角色——这是任何开发者工具（无论是否由AI驱动）或手动操作都可能出现的相同问题。"并指出公司没有收到任何客户关于此次中断的询问。

此外，公司写道："《金融时报》声称有第二起事件影响了AWS，这完全是错误的。"

《金融时报》报道称，亚马逊自己承认了第二起事件，但表示其并未影响"面向客户的AWS服务"。

【更新：文章发布后，一位AWS发言人告诉GeekWire，《金融时报》报道中提及的第二起事件并未发生在AWS业务内部。该发言人表示，《金融时报》的报道在这一点上是错误的。】

至于那起无可争议的影响了AWS的中断事件，《金融时报》的报道引用了四位知情人士的描述，称12月中旬一个AWS系统中断了13小时。

消息人士称，工程师们允许亚马逊的Kiro AI编码工具——一种能够自主采取行动的智能助手——进行更改，而该工具判断最佳行动方案是"删除并重建环境"。

多位亚马逊员工告诉该媒体，这是近几个月来AI工具第二次卷入服务中断事件。根据《金融时报》的报道，一位AWS高级员工表示，这些中断事件"规模虽小但完全可以预见"，并补充说工程师们让人工智能代理在没有人工干预的情况下解决问题。

AWS是亚马逊利润最丰厚的部门。上个季度创造了356亿美元的收入，增长了24%，营业利润为125亿美元。该云部门是公司今年计划2000亿美元资本支出的重点，其中大部分将用于人工智能基础设施。

除了在自身运营中使用自主工具外，亚马逊还在向AWS客户销售这些工具，这使得任何关于AI导致服务中断的叙述都格外不受欢迎。

亚马逊的核心辩护——即12月的事件是"用户错误，而非AI错误"——已经包含在《金融时报》的原始报道中。这篇博客文章在很大程度上以一种更突出、更尖锐的方式重申了这一立场。

亚马逊在回应中写道："我们没有收到任何客户关于此次中断的询问。我们实施了多项保障措施以防止此类事件再次发生——不是因为这次事件影响重大（它并没有），而是因为我们坚持从运营经验中学习，以提高我们的安全性和韧性。"

亚马逊表示，中断仅限于其39个地理区域之一的AWS Cost Explorer服务，这是一款让客户跟踪其云支出的工具。路透社和The Verge援引亚马逊发言人的话报道称，受影响区域位于中国大陆。公司表示，这并未影响计算、存储或数据库等核心服务。

公司补充说，此后已实施了新的保障措施，包括对生产访问进行强制性的同行评审。

《纽约时报》记者迈克·艾萨克在X平台上发帖称，亚马逊的回应是他多年来见过的亚马逊"最带刺的"回应，并将其与公司前公共政策负责人、前白宫新闻秘书杰伊·卡尼曾为其强力辩护的时代相提并论。

英文来源：

Seven hours at the top of Techmeme was apparently too much for Amazon to take.
The tech giant’s cloud division, Amazon Web Services, issued an unusually pointed public rebuttal Friday afternoon to a widely cited Financial Times report asserting that Amazon’s own AI coding tools have caused at least two AWS outages in recent months.
The story was picked up by numerous media outlets, and the widely followed tech news aggregator, as an example of the risks of deploying agentic AI tools, and the underlying question of who — or what — is responsible when something goes wrong.
In a blog post titled “Correcting the Financial Times report about AWS, Kiro, and AI,” Amazon acknowledged a limited disruption to a single service in one region last December but attributed it to a user error in configuring access controls, not a flaw in the AI tool itself.
“The issue stemmed from a misconfigured role—the same issue that could occur with any developer tool (AI powered or not) or manual action,” Amazon said, noting that it received no customer inquiries about the disruption.
In addition, the company wrote, “The Financial Times’ claim that a second event impacted AWS is entirely false.”
The FT reported that Amazon itself acknowledged a second incident but said it did not affect a “customer-facing AWS service.”
[Update: Following publication, an AWS spokesperson told GeekWire that the second event referenced in the FT’s report did not take place within the AWS business. The FT’s reporting is wrong on this point, the spokesperson said.]
As for the undisputed outage impacting AWS, the FT’s report cited four people familiar with the matter in describing a 13-hour interruption to an AWS system in mid-December.
The sources said engineers had allowed Amazon’s Kiro AI coding tool — an agentic assistant capable of taking autonomous actions — to make changes, and that the tool determined the best course of action was to “delete and recreate the environment.”
Multiple Amazon employees told the publication that it was the second time in recent months that AI tools had been involved in a service disruption. According to the FT report, a senior AWS employee said the outages were “small but entirely foreseeable,” adding that engineers had let the AI agent resolve issues without human intervention.
AWS is Amazon’s most profitable division. It generated $35.6 billion in revenue last quarter, up 24%, and $12.5 billion in operating income. The cloud unit is a significant focus of the company’s planned $200-billion capital spending spree this year, much of it directed toward AI infrastructure.
In addition to using agentic tools in its own operations, Amazon is selling them to AWS customers, making any narrative about AI-caused outages particularly unwelcome.
Amazon’s core defense — that the December incident was “user error, not AI error” — was already included in the FT’s original story. The blog post largely restates that position in a more prominent and pointed way.
“We did not receive any customer inquiries regarding the interruption,” Amazon wrote in its response. “We implemented numerous safeguards to prevent this from happening again—not because the event had a big impact (it didn’t), but because we insist on learning from our operational experience to improve our security and resilience.”
Amazon said the disruption was limited to AWS Cost Explorer, a tool that lets customers track their cloud spending, in one of its 39 geographic regions. Reuters and The Verge reported that the affected region was in mainland China, citing an Amazon spokesperson. It did not affect core services such as compute, storage, or databases, the company said.
The company added that it has since implemented new safeguards, including mandatory peer review for production access.
Posting on X, New York Times reporter Mike Isaac called the Amazon response “the most prickly” he’d seen from Amazon in years, comparing it to the past era when former White House press secretary Jay Carney, who led public policy for the company, spoke out strongly in its defense.