11-19 更新 实时
首页/ 科技公司/ 正文

AI 安全失效如何危及你的供应链及应对之策

科技公司 来源:ZDNet 2026-11-10 21:03 0 阅读 机器翻译
AI 安全失效如何危及你的供应链及应对之策

大数据概念。3D渲染图 BlackJack3D/ iStock / Getty Images Plus via Getty Images

令人瞩目的AI应用可能潜藏固有的安全风险。

全球AI公司普遍未能达到基本安全标准。

在采用新AI技术时,供应链监控至关重要。

研究人员警告称,全球人工智能公司正在给数字供应链带来新一波漏洞和安全问题。

AI供应链与安全问题

AI和智能解决方案并不自动等同于安全的解决方案。在一份题为《2026 AI供应链报告:你AI关系背后的风险》的研究中,Black Kite表示,许多AI供应商——即为当今企业开发和部署AI服务的公司——在“基本安全实践”方面存在不足。

根据该研究,在审查了这些公司本身,以及它们的软件、集成和组件之后,研究人员发现它们在修补关键漏洞、保护凭证、验证用户和电子邮件身份,以及解决网络安全问题方面存在失职。

该研究分析了全球18160家AI公司,员工规模从0-10人到1000人以上不等。公司是否涉足AI,是通过对其产品与服务及主营业务的审查来判定的,共涵盖六大类:日常应用开发;AI开发与实施;医疗、IT、金融等专业领域的AI应用;工业组织;以及从事数据、算力和建模的AI基础层机构。

受评估公司中有一小部分被视为“未分类”,该研究还排除了仅偶然使用AI的企业。

最重要的AI供应商风险统计数据

未修复的严重漏洞:严重漏洞并不总会转化为严重的业务风险,但应根据其性质进行风险审查。尽管如此,仍有34.2%(即三分之一)的受评估公司至少存在一个严重漏洞。

电子邮件安全:总体而言,61.7%的AI公司存在DMARC问题,37.3%存在DKIM弱点,二者都可能被滥用于身份冒充和钓鱼攻击。

勒索软件威胁:29.7%的公司极易受到勒索软件攻击。根据该公司的易感性指数,近三分之一的AI供应商所处环境对勒索软件攻击的“暴露程度偏高”。

此外,44.3%的组织被检测到异常网络活动,41.8%的AI公司被记录到与僵尸网络相关的活动。

为什么AI供应商风险值得关注?

AI 的主要商业吸引力在于它能与现有软件和平台集成,处理人工任务、繁重工作和枯燥耗时的杂活,让员工腾出精力专注于需要人来把关的工作。

企业正投入数十亿美元到 AI 上。Menlo 的 2026 年研究估计,2025 年企业 AI 支出达到 370 亿美元,其中一半总支出流向了 AI 应用。总体来看,76% 的企业 AI 解决方案是外购而非自研的。

然而,一旦 AI 或自动化组件被引入企业生态系统或运营平台,它所存在的任何安全问题都会影响用户——而且每一次新增集成,比如软件组件或开源工具,也可能扩大攻击面。

在随附的一份针对 AI 公司使用 Model Context Protocol(MCP)组件的分析中——这些组件可将 AI 应用连接到业务系统——1,732 条记录在案的安全发现中有 48.8% 被评为高风险或严重,其中包括代码注入和命令注入漏洞。

问题不只是 AI 应用本身。正如 Black Kite 所指出的,AI 供应商往往能接触到客户数据、开发者环境、业务基础设施和企业平台。如果供应商的凭证被盗,这可能使其客户面临一系列网络攻击。

例如,研究人员发现,在提供企业 AI 部署解决方案的 AI 公司中,有 19.9% 的公司登录信息出现在窃密软件(infostealer)记录中;而总体而言,被盗的企业登录凭证与研究中 6.8% 的 AI 供应商存在关联。

对于应当监控自身数字供应链的企业而言,可见性同样重要。然而,在接受评估的 AI 公司中,有 33.2% 的数字足迹“有限”,包括其域名、子域名和 IP 地址。

企业应如何应对 AI 供应链风险?

保护数字供应链的相同原则也适用于 AI 供应商,其中持续监控与评估是企业所能拥有的最佳防护手段之一。

及时了解并意识到自身业务面临的风险,会带来截然不同的结果。正如 Black Kite 所建议的,仅靠一次时点性的安全评估远远不够——频繁监控并与供应商沟通、了解其技术和依赖关系,才能揭示潜在的安全问题。

如果发现薄弱环节,应按优先级加以处理,首先处置与勒索软件相关的暴露面,以及已在野被积极利用的漏洞。

此外,还要审计 AI 工具的集成与组件、访问权限、它能触及哪些数据,以及被利用后可能造成的影响。掌握这些信息后,就有可能权衡风险与收益——以及该 AI 工具是否值得承担它所带来的潜在攻击暴露面。如果决定继续保留,就应将其权限和访问范围限制在其必需的范围内。

Charlie Osborne 是一名网络安全记者兼摄影师,在伦敦为 ZDNET 和 CNET 撰稿。PGP 密钥:AF40821B 查看完整简介

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(How AI security failures are endangering your supply chain – and what you can do)

Big data concepts. 3D render BlackJack3D/ iStock / Getty Images Plus via Getty Images

Impressive AI applications can harbor inherent security risks.

AI companies worldwide are failing to reach basic security standards.

Supply chain monitoring is paramount when adopting new AI tech.

Artificial intelligence companies worldwide are bringing a new wave of vulnerabilities and security issues to the digital supply chain, researchers warn.

The AI supply chain and security problems

AI and smart solutions don’t automatically equate with safe solutions. In a study titled “2026 AI Supply Chain Report: The Risks Behind Your AI Relationships,” Black Kite says that many AI suppliers — the companies that develop and deploy AI services to today’s businesses — are struggling with “basic security practices.”

According to the research, after examining the companies themselves, along with their software, integrations, and components, researchers found failures to patch critical vulnerabilities, protect credentials, authenticate users and email, and address network security issues.

The study analyzes 18,160 AI companies worldwide, ranging from 0-10 employees to more than 1,000 employees. A company’s AI involvement was determined through reviews of company products and services and primary business activities, covering six broad groups: everyday application development; AI development and implementation; AI applications in specialized sectors such as healthcare, IT, and finance; industrial organizations; and AI foundations working on data, compute, and modeling.

A small percentage of assessed companies are considered “unclassified,” and the study excluded businesses that use AI only incidentally.

The most important AI supplier risk statistics

Critical, unpatched vulnerabilities: While critical flaws do not always translate to critical business risk, they should be subject to a risk review, depending on their nature. Still, 34.2%, or one in three, of the companies assessed had at least one critical vulnerability.

Email security: In total, 61.7% of AI companies had DMARC issues, while 37.3% had DKIM weaknesses, both of which could be abused for impersonation and phishing.

Ransomware threats: 29.7% of companies were highly susceptible to ransomware. According to the firm’s susceptibility index, almost a third of AI suppliers operate in environments with “heightened exposure” to ransomware attacks.

In addition, unusual network activity was detected at 44.3% of organizations, and botnet-related activity was recorded at 41.8% of AI companies.

Why do AI supplier risks matter?

AI’s main business appeal is its ability to integrate with existing software and platforms to handle manual tasks, heavy lifting, and tedious, time-consuming work, freeing employees to focus on work that requires human oversight.

Enterprises are investing billions of dollars in AI. Menlo’s 2026 research estimates that enterprise AI spend reached $37 billion in 2025, with half of overall spend going to AI apps. In total, 76% of enterprise AI solutions are purchased rather than built internally.

However, the moment an AI or automated component is introduced into a corporate ecosystem or operational platform, any security problems it has impact the user — and each additional integration, such as software components or open-source tooling, may also increase the attack surface.

In an accompanying analysis of AI firms’ use of Model Context Protocol (MCP) components, which can link AI apps to business systems, 48.8% of 1,732 logged security findings were rated High or Critical, including code and command injection vulnerabilities.

It’s not just the AI application itself. As noted by Black Kite, AI suppliers often have access to customer data, developer environments, business infrastructure, and enterprise platforms. If the supplier’s credentials are stolen, this could expose its customers to a range of cyberattacks.

The researchers found, for example, that 19.9% of AI companies offering solutions for enterprise AI deployment had corporate login information in infostealer records, and that stolen corporate logins, overall, were linked to 6.8% of AI suppliers in the study.

Visibility matters, too, for businesses that should be monitoring their digital supply chains. However, 33.2% of AI companies assessed have a “limited” digital footprint, including their domains, subdomains, and IP addresses.

How should businesses respond to AI supply chain risks?

The same principles for protecting the digital supply chain apply to AI suppliers, where continual monitoring and assessment are among the best protections businesses have.

Staying informed and aware of the risks to your own business makes all the difference. As recommended by Black Kite, it takes more than a single point-in-time security assessment — frequent monitoring and engaging with suppliers to understand their technologies and dependencies can reveal potential security issues.

If you find weaknesses, manage them by priority, handling ransomware-related exposures and vulnerabilities actively exploited in the wild first.

Also audit an AI tool’s integrations and components, access, what data it can reach, and the potential impact of exploitation. Once you’re armed with this information, it’s possible to weigh risk and reward — and whether the AI tool is worth the potential attack exposure it brings. If you decide to keep it, limit its permissions and access to what it requires.

Please accept cookies to view the comments powered by Disqus.

Charlie Osborne is a cybersecurity journalist and photographer who writes for ZDNET and CNET from London. PGP Key: AF40821B See full bio

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 ZDNet(How AI security failures are endangering your supply chain – and what you can do),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 ↗