Here’s the Truth About Whether Meta’s NameTag Face Recognition Tech ‘Exists’

关于 Meta 的 NameTag 人脸识别技术是否“存在”，真相究竟如何

Does a software feature exist if its code has been deployed to the devices of millions of people but they can’t use it yet? Not if you work at Meta. 如果一个软件功能已经部署到数百万人的设备上，但用户还无法使用它，那么这个功能算存在吗？如果你在 Meta 工作，答案是否定的。

The company’s executives have spent the last few weeks making this semantic argument about NameTag, the in-development face-recognition system that Meta built for its smart glasses. The inevitable result is confusion, but that’s easy enough to clear up. 过去几周，Meta 的高管们一直在围绕 NameTag 进行这种语义上的辩论。NameTag 是 Meta 为其智能眼镜开发的人脸识别系统。这种辩论不可避免地导致了困惑，但要理清这一点并不难。

On June 4, WIRED reported that Meta included robust—but inactive—code for NameTag in Meta AI, the companion app for Meta Ray-Ban glasses that has been downloaded tens of millions of times. In response to our story, Andy Stone, Meta’s vice president of communications, responded in part by writing on X, “Here’s a thing: Wired reports Meta didn’t answer several questions about how this will work. How could we? The feature doesn’t exist!” Meta removed the NameTag code from Meta AI the next day. 6 月 4 日，《连线》（WIRED）报道称，Meta 在 Meta AI 中植入了强大但处于非激活状态的 NameTag 代码。Meta AI 是 Meta Ray-Ban 智能眼镜的配套应用，下载量已达数千万次。针对我们的报道，Meta 通讯副总裁安迪·斯通（Andy Stone）在 X 上回应称：“有一点：WIRED 报道称 Meta 没有回答关于该功能如何运作的几个问题。我们怎么回答？这个功能根本不存在！”次日，Meta 从 Meta AI 中移除了 NameTag 代码。

According to WIRED’s analysis of the Meta AI app, code for NameTag appeared in the app as early as January. In mid-February, The New York Times reported that Meta had been working on NameTag face recognition. By May, WIRED found, the core components of the NameTag code were present in the MetaAI app. 根据《连线》对 Meta AI 应用的分析，NameTag 的代码早在 1 月份就已出现在应用中。2 月中旬，《纽约时报》报道称 Meta 一直在研发 NameTag 人脸识别技术。到了 5 月，《连线》发现 NameTag 代码的核心组件已经存在于 Meta AI 应用中。

Whether the feature existed prior to Meta removing the code depends on how one defines “feature” and “exist.” Whatever one’s position, a researcher who goes by the name Buchodi reviewed the code at WIRED’s request, and was able to use the NameTag system to recognize a photograph of the face of the philosopher Michel Foucault, famed for his writing on surveillance as an instrument of power. 在 Meta 移除代码之前，该功能是否“存在”，取决于人们如何定义“功能”和“存在”。无论立场如何，一位名为 Buchodi 的研究人员应《连线》要求审查了这些代码，并成功利用 NameTag 系统识别出了哲学家米歇尔·福柯（Michel Foucault）的照片，福柯以撰写关于监控作为权力工具的著作而闻名。

The claim that Meta had no way of describing how the feature works—or even would work—was further undercut last week, when Meta CTO Andrew “Boz” Bosworth did so in detail on a podcast. Meta 声称无法描述该功能如何运作（甚至是如何“将要”运作）的说法，在上周进一步被推翻，因为 Meta 首席技术官安德鲁·“博兹”·博斯沃思（Andrew “Boz” Bosworth）在一个播客节目中详细描述了其运作方式。

On the July 8 episode of the The Most Interesting Thing in AI, host Nicholas Thompson—CEO of The Atlantic and a former WIRED editor in chief—asked who NameTag would identify during a part of the discussion labeled “What’s true and false about NameTag.” Bosworth replied: “Somebody you met in person with your glasses on who introduced themselves—or you said, ‘OK, this is David, remember this person.’ Only available to you when you’re wearing your glasses—this is a person you’ve met before. Here’s their name. They’re right in front of you … That’s what we call a NameTags feature.” 在 7 月 8 日播出的《人工智能中最有趣的事》（The Most Interesting Thing in AI）节目中，主持人尼古拉斯·汤普森（Nicholas Thompson，现任《大西洋月刊》CEO，曾任《连线》主编）在讨论“关于 NameTag 的真与假”环节中，询问 NameTag 会识别谁。博斯沃思回答道：“是你戴着眼镜时亲自见过并进行过自我介绍的人，或者你说过‘好的，这是大卫，记住这个人’。只有当你戴着眼镜时才可用——这是你以前见过的人。这是他们的名字。他们就在你面前……这就是我们所说的 NameTags 功能。”

Bosworth later said of NameTag, “So, it’s a thing that, um—I think would be a great feature.” 博斯沃思随后评价 NameTag 说：“所以，它是一个……嗯，我认为会是一个很棒的功能。”

In response to inquiries from WIRED about this apparent contradiction, Meta has repeatedly stressed the conditional nature of Bosworth’s statement—that it “would be” a great feature, not that it is or will be. Spokesperson Ryan Daniels specifically highlighted the word “would,” bolding and underlining it, in an email exchange with WIRED about the apparent disconnect between Stone’s claim that NameTag doesn’t exist and Bosworth’s minutes-long description of it. 针对《连线》关于这一明显矛盾的质询，Meta 反复强调博斯沃思言论中的条件性——即它“会是”（would be）一个很棒的功能，而不是它“是”或“将是”一个功能。在与《连线》的邮件往来中，发言人瑞安·丹尼尔斯（Ryan Daniels）特别强调了“would”这个词，并对其进行了加粗和下划线标注，以回应斯通声称 NameTag 不存在与博斯沃思长达数分钟的描述之间明显的脱节。

“There is no contradiction. Boz says this ‘would’ be a good feature, particularly to answer the blind and low-vision community members’ calls to help them identify people they’ve already met or want to remember,” Daniels tells WIRED in a statement. “While we’re exploring this, it’s not available to consumers today. We think it’s important that people understand this remains distinct from connecting glasses to a central database of people in the world, which is not a capability we are building.” “这并不矛盾。博兹说这‘会是’一个很好的功能，特别是为了响应盲人和低视力群体成员的呼吁，帮助他们识别已经见过或想要记住的人，”丹尼尔斯在给《连线》的声明中表示。“虽然我们正在探索这一点，但它目前并未向消费者开放。我们认为让人们理解这一点很重要：这与将眼镜连接到全球中央人脸数据库有着本质区别，而后者并非我们正在构建的功能。”

To be clear, NameTag existed as of roughly six weeks ago. Meta had been building NameTag since early 2025—licensing face-recognition software, assembling a full detection-and-matching pipeline, and adding it to the tens of millions of phones the app runs on, where it sat until WIRED reported on it. While it was not possible for people to actually use it without specialized tools, WIRED’s analysis of the Meta AI code, as well as that of two independent experts, found a technically functional face-recognition system within the app millions of people have on their phones. That system still exists, if you take Bosworth’s discussion of it at face value. 明确一点，大约六周前 NameTag 是存在的。Meta 自 2025 年初以来一直在构建 NameTag——授权人脸识别软件、组装完整的检测与匹配流程，并将其添加到数千万部运行该应用的手机中，直到《连线》报道此事之前，它一直存在于那里。虽然人们在没有专业工具的情况下无法实际使用它，但《连线》对 Meta AI 代码的分析，以及两位独立专家的分析都发现，在数百万人的手机应用中存在一个技术上可运行的人脸识别系统。如果你从字面上理解博斯沃思的讨论，那么该系统依然存在。

Even so, Meta has insisted, and continues to insist, that that’s not quite true, in ways it declines to specify. Despite WIRED stating in the first full sentence of the June 4 article that the NameTag system was “unreleased” and reiterating that point repeatedly throughout, for example, Stone claimed on X that the story made this detail unclear. “This is more than shoddy reporting, it’s intellectually dishonest,” Stone wrote. “Pure advocacy-driven click bait.” Bosworth responded to Stone’s X post by calling WIRED’s report “incredibly misleading” and “absolutely dishonest.” 即便如此，Meta 仍然坚持认为事实并非如此，但拒绝说明具体原因。例如，尽管《连线》在 6 月 4 日文章的第一句完整句子中就明确指出 NameTag 系统是“未发布”的，并在全文中反复重申了这一点，但斯通仍在 X 上声称该报道使这一细节变得模糊。“这不仅仅是粗制滥造的报道，更是智力上的不诚实，”斯通写道，“纯粹是受立场驱动的点击诱饵。”博斯沃思回应了斯通的 X 帖子，称《连线》的报道“极具误导性”且“绝对不诚实”。

Meta did not answer WIRED’s questions about what its executives believed was misleading or dishonest about our NameTag reporting. 对于其高管认为我们的 NameTag 报道中哪些部分具有误导性或不诚实，Meta 并未回答《连线》的提问。

One point is clearly important to Meta: Bosworth asserted in his conversation with Thompson that NameTag will not rely on a “central database,” even though neither Thompson nor WIRED claimed or suggested it would. Instead, WIRED’s analysis of the app found, the NameTag system converts faces captured by Meta glasses and turns them into unique numerical signatures commonly known as “faceprints.” These faceprints could be compared against a face database stored on users’ devices that was populated by Meta’s servers. 有一点对 Meta 显然很重要：博斯沃思在与汤普森的谈话中坚称 NameTag 不会依赖“中央数据库”，尽管汤普森和《连线》从未声称或暗示过它会这样做。《连线》对该应用的分析发现，NameTag 系统会将 Meta 眼镜捕捉到的人脸转换为独特的数字签名，通常被称为“人脸指纹”（faceprints）。这些“人脸指纹”可以与存储在用户设备上、由 Meta 服务器填充的人脸数据库进行比对。

The distinction between a “central” database and millions of local databases (i.e., your phone and mine) connected to Meta’s servers is not merely academic. As Bosworth mentioned during his discussion with Thompson, some state laws, such as the Biometric Information Privacy Act (BIPA) in Illinois and Texas’ Capture or Use of Biometric Identifier Act (CUBI), limit companies’ ability to publicly deploy face-recognition technology, which require people to explicitly consent to having their faces captured by face-recognition systems, among other guardrails. “中央”数据库与连接到 Meta 服务器的数百万个本地数据库（即你我的手机）之间的区别不仅仅是学术上的。正如博斯沃思在与汤普森讨论时提到的，一些州法律，例如伊利诺伊州的《生物识别信息隐私法》（BIPA）和德克萨斯州的《生物识别标识符捕获或使用法》（CUBI），限制了公司公开部署人脸识别技术的能力，这些法律要求人们明确同意其面部被系统捕获，并设有其他防护措施。

Bosworth would know. In 2019, Meta abandoned its “Tag Suggestions” automatic face-recognition feature on Facebook af 博斯沃思对此深有体会。2019 年，Meta 放弃了其在 Facebook 上的“标签建议”（Tag Suggestions）自动人脸识别功能。