Eightfold 遭遇集体诉讼：一场所有北美 HR 都必须看懂的 AI 招聘警示2026 年 1 月，美国 AI 招聘平台 Eightfold AI 在加州州法院遭遇拟议集体诉讼。两名求职者指控 Eightfold 在招聘流程中，未经候选人知情或授权，生成并提供用于筛选的候选人评估报告，涉嫌违反《公平信用报告法》（Fair Credit Reporting Act，FCRA）以及加州消费者保护相关法律。这起案件被多家法律机构视为美国首批直接以 FCRA 为依据起诉 AI 招聘平台的集体诉讼之一，其影响已经超出单一厂商范畴，直指整个 AI 招聘生态。
对许多 HR 而言，这起诉讼之所以引发困惑，是因为 Eightfold 并不是传统意义上的背景调查公司，也并未直接作出雇佣决定。那么，问题究竟出在哪里？
Eightfold 被指控“做了什么”？
根据起诉文件，Eightfold 为企业客户提供基于 AI 的招聘与人才匹配工具，系统会综合分析候选人的简历、职位描述和历史职业信息，自动生成所谓的“人才画像”。这些画像包含对候选人性格特征（如“团队型”“内向型”）、教育质量、岗位匹配度以及未来职业路径的预测，并被企业用于招聘筛选和排序。
据代表原告的律师事务所之一Outten & Golden LLP称，两名原告提起诉讼的原因是他们认为自己被排除在符合条件的职位候选名单之外。
原告的核心指控并不是算法“算错了人”，而是认为 Eightfold 在法律意义上扮演了第三方评估者的角色：其系统性生成的评估结果被用于影响雇佣决策，但候选人并未被明确告知这些评估的存在，也无法查看、质疑或纠正其中可能存在的错误。
为什么会牵扯到 FCRA？
在北美 HR 语境中，FCRA 通常被理解为一部规范信用报告和背景调查的法律。但法律的关键并不在于“是否查询了征信”，而在于是否存在第三方生成的、用于雇佣决策的个人评估报告。
原告的法律逻辑是：如果 Eightfold 生成的 AI 评估结果在事实层面等同于“消费者报告”，那么候选人就应当享有法律赋予的基本程序性权利，包括事前通知、查看报告以及纠错的机会。Eightfold 是否使用 AI、是否出于善意，并不是争议焦点；焦点在于招聘判断是否被规模化外包给一个候选人完全不可见的系统。原告反复强调“候选人完全不知情”，而不是“未同意”，这样的表达是告诉你：候选人 根本不知道 自己正在被一个第三方系统评估，更不知道这些评估会被用于筛选或排序。
很多人第一反应是把这起案件理解为“AI 歧视诉讼”，我们研读起诉书后发现，原告是刻意没有走这条路的。这是一个非常专业且稳妥的律师！原告把全部火力集中在一个更“稳”的法律切口上：
Eightfold 在法律上是否构成 Consumer Reporting Agency，以及是否履行了 FCRA 要求的程序性义务。
Eightfold 的回应与案件现状
Eightfold 对外回应称，其平台使用的数据来自候选人本人或客户提供，并不抓取社交媒体等外部信息，同时强调其对负责任 AI 和合规的承诺。目前，该案件仍处于司法程序早期阶段，法院尚未就实体问题作出裁决。
但对 HR 而言，真正重要的并非案件最终胜负，而是法院已经允许这一诉讼路径进入审理轨道。这意味着，AI 招聘平台是否可能被纳入现有消费者保护法律的监管范围，已经不再只是理论讨论。
这对北美 HR 意味着什么？
这起案件释放出的信号，并不是“HR 不该使用 AI”，而是：使用 AI 招聘工具并不等于风险外包。在美国法律体系下，雇主与技术供应商之间的责任边界，正在被重新审视。
对于 HR 来说，一个常见的误区是认为“最终是否录用由人决定，因此 AI 不需要承担程序性责任”。但司法系统正在关注的是：当候选人筛选高度依赖第三方系统时，雇主是否仍然可以完全享有“人类裁量权”的豁免。1月北美华人HR洛杉矶新年论坛中，就有嘉宾特别谈到了在HR工作中使用负责任的AI，也谈到了部分AI招聘服务提供商违规使用面试者的面试内容作为商业竞争的信息来源引发的争议。
NACSHR特别的提醒
需要强调的是，目前诉讼的焦点仍然集中在 Eightfold AI 这一技术提供商身上，并不意味着雇主已经被排除在风险之外。更准确地说，这是司法体系在处理新型技术风险时的一种常见路径——先审视系统性、可复制的技术行为是否合法，再讨论使用这些系统的雇主应当承担何种责任。对雇主而言，当前阶段的风险并非“已经发生”，而是“正在被重新界定”。在 AI 招聘被逐步纳入既有法律框架的过程中，继续将责任完全视为供应商问题的空间正在缩小。真正理性的应对方式，并不是等待雇主成为下一个被告，而是在这一窗口期内，主动理解 AI 系统如何影响招聘决策，并提前将其纳入企业的合规与治理视野之中。
从起诉书本身来看，原告的策略并非指控算法歧视或技术失误，而是刻意将争议锚定在程序性合法性之上。通过反复强调 Eightfold 对候选人性格、潜力和未来路径的预测性评估，以及候选人对这些评估完全不知情的事实，起诉书试图把 AI 招聘系统重新定义为一种受监管的第三方评估机制，而非中立工具。这种策略的真正锋芒不在于推翻某一产品，而在于为整个 AI 招聘行业设定新的法律边界。
NACSHR给北美 HR 的现实建议（可操作）
第一，重新审视你正在使用的 AI 招聘工具的角色定位。HR 需要明确：当前系统是在“辅助判断”，还是事实上已经在自动排序、淘汰候选人？是否存在候选人因为 AI 评分而从未进入人工视野的情况？这是风险评估的第一步。
第二，向供应商明确询问合规与候选人权利设计。HR 应主动询问以下问题：候选人是否被告知 AI 的使用？是否可以查看与自身相关的评估输出？是否存在纠错或申诉机制？如果这些问题无法得到清晰回答，企业就有必要重新评估使用方式。
第三，不要将“解释义务”与“给拒绝理由”混为一谈。法律并未要求 HR 为每一次拒绝提供主观理由，但当评估被外包给第三方系统时，程序性透明和最低限度的候选人权利保障，正在成为不可忽视的合规要求。
第四，让 Legal 和 Compliance 参与 AI 招聘决策。AI 招聘已经不再是单纯的 HR 技术选型问题，而是潜在的法律和声誉风险源。将其纳入企业合规框架，而不是仅作为效率工具使用，是当前北美 HR 的理性选择。
Eightfold 的集体诉讼并不是一次偶然事件，还记得Workday因为AI招聘的诉讼案吗？而是北美司法系统对 AI 招聘现实影响的集中回应。对 HR 来说，这并不意味着要放弃技术，而是意味着必须更清醒地理解：当招聘判断被系统化、规模化并外包给 AI 时，企业对候选人所承担的责任，并不会因此消失。未来真正可持续的 AI 招聘实践，将取决于 HR 是否能够在效率、合规与人类裁量之间，找到新的平衡点。
你想知道Eightfold 可能早都发现了这个可能的风险，他们做了你想不到的一步棋，你猜猜是什么？欢迎留言！
Eightfold
2026年01月24日
Eightfold
Josh Bersin: When Will The Trillions Invested In AI Pay Off? Sooner Than You Think.近年来，生成式人工智能（GenAI）的投资已达数万亿美元，但围绕其回报问题的争论不断升级。一些分析师，如麻省理工学院教授达隆·阿西莫格鲁（Daron Acemoglu）和纽约大学心理学与神经科学教授加里·马库斯（Gary Marcus），对AI的经济影响持悲观态度，认为其对美国生产力和GDP增长的推动作用有限，甚至可能导致市场崩溃。相反，另一派如高盛的全球经济学家则乐观地认为，AI有望在未来十年内大幅提高生产力。然而，文章指出，生成式AI的真正价值在于其特定领域的应用。例如，Paradox和Galileo等HR技术平台通过高度专业化的解决方案，显著提升了招聘和人才管理的效率。最终，文章强调，AI行业仍处于早期阶段，成功的关键在于找到具有专注性和精确性的创新解决方案。
In the last few weeks there has been a lot of concern that Gen AI is a “bubble” and companies may never see the return on the $Trillion being spent on infrastructure. Let me cite four analyst’s opinions.
Will Today’s Massive AI Investments Pay Off?
MIT professor Daron Acemoglu estimates that over the next ten years AI will impact less than 5% of all tasks, concluding that AI will only increase US productivity by .5% and GDP growth by .9% over the next decade. As he puts it, the impact of AI is not “a law of nature.”
On a similar vein, Gary Marcus, professor emeritus of psychology and neural science at New York University, believes Gen AI is soon to collapse, and the trillions spent will largely result in a loss of privacy, increase in cyber terror, and a lack of differentiation between providers. The result: a market with low profits and big losses.
Goldman Sachs Head of Equity Research Jim Covello is similarly pessimistic, arguing simply that the $1 Trillion spent on AI is focused on tech that cannot truly automate complex tasks, and that vendors’ over-focus on “human-like features” will miss the boat in delivering business productivity. (He studies stocks, not the economy.)
And Goldman Sachs Global Economist, who is a fan, estimates that AI could automate 25% of work tasks and raise US productivity by 9T and GDP by 6.1% over the next decade. He follows the traditional business meme that “AI changes everything” for the better.
What’s going on? Quite simply this new technology is very expensive to build, so we’re all unsure where the payoffs will be.
Buyers Are Looking For A Return Soon
If we discount the work going on at Google, Meta, Perplexity, and Microsoft to build AI-based search businesses, which make money on advertising (Zuckerberg essentially just said that in a few years AI will guarantee your ad spend pays off), corporate IT managers are asking questions.
An article in Business Insider pointed to a large Pharma company that cancelled their Microsoft Copilot licenses because the tool was not adding any significant value (Chevron’s CIO was quoted similarly in The Information).
Another quoted a Chief Marketing Officer who stated Google Gemini’s email marketing tool and the new AI-powered ad-buying tool performed worse than the human workers it was intended to replace (or support).
Given that these tools almost double the “price per user” for the productivity suites, I think it’s fair that CIOs, CMOs, to expect them to pay for themselves fairly quickly.
What’s Going On? The Big Wins Will Be Domain Specific
As with all new technologies that enter the market quickly, “the blush on the rose” is over. We’ve been dazzled by the power of ChatGPT and now we’re searching for real solutions to problems. And unlike the internet, where research was funded by the government, there’s going to be a lag (and some risk) between the trillions we spend and the trillions we save.
Given that ChatGPT is less than two years old and OpenAI has morphed from a research company into a product company, it’s easy to see what’s happening. Every vendor and tool provider is narrowing its AI “strategy” and not just pasting little AI “stars” on their websites, looking for useful things to do. And this process may take a few years.
In the world of HR, I think we can all agree that a “push the button job description generator” is a bit of a commodity. However if the AI analyzes the job title, identifies the skills needed through a large skills engine, and tunes the job description by company size, industry, and role, then it’s a fantastic solution. (Galileo does this, as does SeekOut, SAP, and some other vendors.)
The more “specific” and “narrow” the AI is, the more useful it becomes. Generic LLMs that aren’t highly trained, optimized, and tuned to your company, business, and job are simply not going to command high prices. So while we all thought ChatGPT was Nirvana, we’re now figuring out that highly specialized solutions are the answer.
Let me give you some examples.
The first is the platform built by Paradox, a pioneering company that started work on AI-based recruiting agents in 2016. Paradox, now valued at around $2 Billion, delivers an end-to-end recruitment platform that automates the entire process of candidate marketing, candidate experience, assessment, selection, interview scheduling, hiring, and onboarding. Most people believe its a “Chatbot” but in reality it’s an AI-powered end-to-end system that radically simplifies and speeds the recruitment process in a groundbreaking way. Companies like 7-11, FedEx, GM, and others see massive improvements in operational efficiency and both candidates, managers, and recruiter adore it. It took Paradox eight years to build this level of integrated solution.
The second is our platform Galileo. Galileo, which is now licensed by more than 10,000 HR professionals, is a highly tuned AI agent specifically designed to help HR professionals (leaders, business partners, consultants, recruiters, and other roles) do the “complex work” HR professionals do. It’s not a generic LLM: it’s a highly specialized solution designed specifically for HR professionals, and we’ve added specialized content partners and are building special integrations with other HR platforms. Our clients tell us it’s saving them 1-2 hours a day.
The third is the platform HiredScore, that was recently acquired by Workday. Founded in 2012, the HiredScore team built tools to help identify “fit” between individuals and jobs, and tuned its AI to be highly explainable, unbiased, and very easy to use. It took Athena Karp and the team a few years to nail down the use-cases and user interface but now HiredScore is considered one of the most powerful recruitment “orchestration” tools in the market, and is also used for internal hiring and many other applications. Every customer I talk with tells me it’s essential and saves them months of manual, error-prone effort.
The fourth is the platform Eightfold, which was invented in 2016 as a way to build “Google-scale” matching between job seekers and jobs. Through many years of engineering, product management, and ongoing sales process the company has become the leader in a new space called “Talent Intelligence,” now a billion dollar rapid-growing category. The company is about ten years old and now has some of the world’s largest companies building their hiring, career management, and talent management processes using AI. Companies like EY, Bayer, and Chevron now use it for all their strategic talent programs.
Each of these vendors, including others like Gloat, Sana, Arist, Lightcast, Draup, Uplimit, Firstup, and hundreds of others have patiently taken the power of Generative AI and applied it with laser precision to their solutions. Each of these companies is different, and as we work with them we see lightning bolts of innovation: not in AI itself, but in finding new ways to solve problems and do what I call “crawling up the value curve.”
This is the path for AI in the coming years. As with all new technologies, the “trough of disappointment” is always followed by the “bowling pin” of hitting the nail on the head. Innovators, entrepreneurs, and startup founders are the ones who will take GenAI and apply it in unique ways to solve problems. And soon enough, “AI-powered” will be a phrase we barely even need to say.
The Best Solutions Will Be Narrow Not Wide
GenAI solutions require a large “platform” of data, infrastructure, and software. That alone is not where the value resides. Rather, the big productivity advantages come after years of effort, focusing the data sets and working with customers to find the features, UI designs, and data sets that add enormous value. And we are still in the early stages.
If you want to learn more about HR Technology and AI, join me at the HR Technology Conference on September 24-25 in Vegas, or at Unleash in Paris in October 16-17. While I can’t predict who will win the core AI platform game (Microsoft, OpenAI, Google, Meta, Amazon will fight it out), I can predicts this: Generative AI will deliver massive improvements in business productivity. You just have to shop around a bit and wait for just the right solutions to arrive.
Eightfold
Eightfold
扫一扫
添加客服
