11-20 更新 实时
首页/ 公司动向/ 正文

这些技术故障颇有千年虫味道,AI 智能体版

公司动向 来源:TechCrunch AI 2026-11-20 05:42 0 阅读 机器翻译
这些技术故障颇有千年虫味道,AI 智能体版

正如美国说唱歌手 Yung Joc 可能会说的那样,它正在崩掉:航空公司订票系统、银行平台和生产力平台。过去两周,美国顶级企业接连遭遇技术故障,扰乱了支撑日常生活的各类系统。

Apple Pay 宕机。NASA 的内部通信网络中断。FAA 不得不让所有航班停飞数小时(这是自 9/11 以来历史上仅有的第三次),而 Microsoft 遭到攻击,导致 Outlook、Teams 和 Copilot 瘫痪了近两天。

第一反应自然是某种形式的 AI 黑客入侵,尤其是在几周前网络黑客成功攻击了一家杂货店的冰箱和法国一家血库之后。在美国,国家情报总监表示,这些故障似乎与国家支持的恐怖主义无关,而 Trump 总统则称这些问题是“孤立事件”。许多公司自己也表示,故障并非由网络攻击或 AI 攻击造成,而是技术问题。

只不过,这些故障全都发生在彼此相隔很短的时间窗口内,而且就在网络安全界所谓的“黑色十月”几周之后——当时在 10 月 21 日也发生了一连串同样互不相关的网络安全事件,波及 PHP、Node.js、LiteLLM 等。

这看起来确实不像是巧合。坦白说,这给人一种“AI 原生版千年虫”的感觉。

Dave DeWalt 是一位在网络安全领域工作了三十多年的投资人(曾担任 McAfee 总裁兼 CEO,以及美国国土安全部 CISA 咨询委员会副主席),他也认为这很可能与 AI 有关。

他告诉 TechCrunch,这些技术故障很可能是“一个正在变强而非变弱的安全系统所经历的成长阵痛”,而且“AI 正通过 Mythos 等 AI 工具,以前所未有的速度和规模暴露漏洞”。

问题在于,这些技术缺陷中有许多一直存在,但由于 AI 暴露它们的速度太快,企业正抢在恶意行为者加以利用之前紧急修复。“这波补丁潮短期内打乱了节奏,”他说。

软件工程师、安全初创公司 Corsha 创始人 Anusha Iyer 告诉 TechCrunch,这或许是传统与现代交汇的第一个时刻。“这些工业系统从来就不是为联网而设计的,更不用说以我们今天推动的方式安全联网了,”她说。

弗吉尼亚州参议员、参议院情报委员会副主席 Mark Warner 得出了类似的结论(尽管针对可能发生的情况,调查仍在进行中)。“这些中断应当敲响警钟。感觉每天都有新的提醒,说明 AI 的发展速度已经超出了我们管理其后果的能力,”这位参议员在 X 上写道。

AI 正在重塑软件的构建、部署和运行方式,其规模之大让每个人都感受到陡峭的学习曲线。Iyer 以 FAA 的故障为例,说明传统技术并不像企业试图与之融合的那些技术那样精密或先进。

用不那么技术化的话说,她表示这次故障的起因可能是(初步调查结果也指向这一点):FAA 当时正试图对其电信基础设施进行现代化改造,而这一过渡压垮了其部分遗留硬件——新技术很可能以正常的现代容量运行,而遗留硬件却跟不上。

她说,从这个角度看,AI 其实背了黑锅。“目前导致这些故障的并不是 AI 的快速采用,”她说。“问题在于遗留系统与现代系统被仓促混用,却对运营风险和网络风险缺乏清晰认识,也没有把关和保护遗留系统的能力。”

DeWalt 表示,各机构需要以更稳健的方式采用 AI,包括进行更多测试,并确定哪些问题需要修复、何时修复的优先级。

Iyer 补充道:“这些工业系统需要更严格地隔离、更严密地控制访问权限,这样在引入现代系统时,就能分阶段推进并降低风险。”

实际上,对相关公司(以及风险投资者)来说,这一切或许还有一线希望。DeWalt 表示,他宁愿看到这些美国企业以这种混乱的方式发现并修补自身缺陷,也不愿看到恶意行为者借 AI 战找到并利用这些漏洞,从而引发更严重的国家安全问题。

也许一些有抱负的创业者还会从中受到启发。DeWalt 预计,专注于“AI 驱动的漏洞管理、自主修复、安全补丁测试与部署,以及维持关键基础设施运行的系统韧性”的公司将会增多,他说。

“正如我们过去一周所见,这一转型确实带来了切实的痛苦,”DeWalt 说。“但对于那些帮助组织以 AI 速度前进而不出乱子的公司来说,也存在着巨大的机遇。”

当你通过我们文章中的链接进行购买时,我们可能会赚取少量佣金。这不会影响我们编辑的独立性。

资深记者,风险投资

Dominic-Madori Davis 是 TechCrunch 的资深风险投资与初创企业记者。她常驻纽约市。

你可以发送邮件至 [email protected],或通过 Signal 上的 +1 646 831-7565 发送加密信息,来联系 Dominic 或核实其联系请求。

正在加载下一篇文章

译文由英文原文机器翻译,术语按行业惯例统一,可在文末展开原文核对。

查看英文原文(These tech outages are giving Y2K, AI agent style)

It’s going down, as the American rapper Yung Joc might say: the airline booking systems, banking platforms, and productivity platforms. In the past two weeks, the U.S.’s top businesses have faced technical outages that have disrupted the systems that run everyday life.

Apple Pay went down. NASA’s internal communication network went down. The FAA had to ground all flights for hours (only the third time in history since 9/11), and Microsoft was hit with an attack that took down Outlook, Teams, and Copilot for nearly two days.

The first thought, naturally, is an AI hacker invasion of some sort, especially after cyber hackers managed to attack fridges in a grocery store and a blood bank in France a few weeks ago. Here in the U.S., the Director of National Intelligence said these outages don’t seem related to state-sponsored terrorism, while President Trump called the issues “individual events.” Many of the companies themselves have said the outages were not caused by cyberattacks or AI attacks, but tech issues.

Only, there’s the fact that they all struck within a short timespan of each other, and just a few weeks after Cybersecurity’s so-called “Black October,” when a slew of also-unrelated cybersecurity incidents occurred on October 21 striking PHP, Node.js, LiteLLM, more.

It does seem like more than a coincidence. It’s giving AI-native Y2K, frankly.

Dave DeWalt, an investor who worked in cybersecurity for more than three decades (including as president and CEO of McAfee and as vice Chairman of the CISA Advisory Council for the Department of Homeland Security), also thinks it’s probably AI-related.

The technical outages are most likely the “growing pains of a security system that’s getting stronger, not weaker,” and that “AI is surfacing vulnerabilities at a speed and scale we’ve never had before through AI tools like Mythos and others,” he told TechCrunch.

The thing is, many of these technical flaws have always existed, but because AI is surfacing them so fast, enterprises are rushing to fix them before bad actors can exploit them. “That surge in patching is disrupting things in the short term,” he said.

This was, perhaps, the first moment when legacy meets modern, Anusha Iyer told TechCrunch, a software engineer and founder of the security startup Corsha. “These industrial systems were never designed to be connected, let alone securely connected in the way we are pushing them to be today,” she said.

Virginia Senator Mark Warner, vice chair of the Senate Intel Committee, came to a similar conclusion (although investigations are still underway as to what could be happening). “These disruptions should serve as a wake-up call. It feels like every day brings another reminder that AI is moving faster than our ability to manage its consequences,” the Senator wrote on X.

AI is reshaping how software is built, deployed, and operated, on a scale where everyone feels the sharp learning curve. Iyer cited the FAA outage as an example of legacy tech that is not as sophisticated or advanced as the technologies companies are trying to merge it with.

In less technical terms, she said the outage may have been caused (and preliminary findings point to this) when the FAA was trying to modernize its telecommunications infrastructure and the transition overwhelmed parts of its legacy hardware, where the new tech likely operated at a normal modern capacity while the legacy hardware couldn’t keep up.

When looked at this way, she said, AI is actually getting the bad rep. “It’s not rapid adoption of AI yet causing these outages,” she said. “The fault is the rushed mix of the legacy and modern systems without a clear understanding of operational and cyber risk or the ability to gatekeep and protect the legacy systems.”

Organizations need a more steadfast approach to AI adoption, including more testing and prioritizing what needs to be fixed and when, DeWalt said.

Iyer added, “These industrial systems need to be segmented more tightly and more access controlled, so as modern systems are introduced, it can be phased in and risk mitigated.”

Actually, there might be a silver lining to all of this for companies (and for venture investors). DeWalt said he’d rather see these U.S. enterprises find and patch their flaws, even in this messy way, than have bad actors find and exploit those vulnerabilities in AI warfare that could cause even more national security issues.

And maybe some aspiring founders were inspired by them. DeWalt expects a rise in companies focused on “AI-driven vulnerability management, autonomous remediation, safe patch testing and deployment, and resilience for the systems that keep critical infrastructure running,” he said.

“There’s real pain in this transition as we’ve seen in the past week,” DeWalt said. “But there’s also enormous opportunity for the companies that help organizations move at AI speed without breaking things.”

When you purchase through links in our articles, we may earn a small commission. This doesn’t affect our editorial independence.

Dominic-Madori Davis

Senior Reporter, Venture

Dominic-Madori Davis is a senior venture capital and startup reporter at TechCrunch. She is based in New York City.

You can contact or verify outreach from Dominic by emailing [email protected] or via encrypted message at +1 646 831-7565 on Signal.

Loading the next article

原文为英文,正文由自动抓取并保留原文。

本文由采集器自动同步

原文发布于 TechCrunch AI(These tech outages are giving Y2K, AI agent style),标题、摘要与正文由机器翻译,术语按行业惯例统一,未经人工逐句校订。

阅读原文 ↗