2026-09-30T14:07:00-0400 / 哥伦比亚广播公司新闻
面对近期外界对其技术隐私成本的质疑，监控摄像头初创公司Flock Safety及其首席执行官抛出了一项简洁的统计数据来宣扬其技术益处。
“去年我们帮助美国破获了刚过100万起案件，”Flock首席执行官加勒特·兰利在2026年3月的播客采访中说道。
“去年，这项技术帮助破获了超过100万起案件，”兰利在接受汽车行业出版物《The Drive》采访时表示，“相比之下，滥用情况极低。”
“如今，Flock技术每年协助执法人员破获超过100万起案件，”兰利在上个月宣布新隐私更新的一篇博客文章中说道。
但哥伦比亚广播公司新闻的调查发现，Flock称其在2025年协助破获100万起案件的说法，并无现有证据支持。
在为本报道提供的声明中，Flock发言人帕里斯·卢贝尔对这一统计数据给出了不同的解释。他表示，除了案件告破之外，该数据还包括“调查线索、跨司法管辖区协助、主动调查期间开展的排查以及其他Flock部署应用场景”。
“描述我们2025年全国估算数据最准确的方式是，Flock技术全年支持了约100万起调查和事件，”卢贝尔说道。
该公司未回应兰利称Flock在2025年破获100万起案件的说法是否不够精准，也未向哥伦比亚广播公司新闻提供支持该说法的原始数据。
Flock最早在2026年5月的一篇博客文章中解释了其“破获100万起案件”说法的来源，该文章公布了其2025年影响力普查的结果，约700家使用Flock系统的执法机构参与了此次普查。Flock表示，正式追踪办案结果的机构提供了原始数据，而其他机构则提供了百分比估算值。该博客文章还称，研究发现2025年客户辖区内20%的已告破案件得到了Flock技术的协助。
在解释这20%的调查结果时，该公司表示有307家机构提供了百分比估算值。随后这些回应被与每个响应机构的FBI犯罪和案件告破数据进行了关联。
“100万起案件”的统计数据是一项独立且范围更广的计算结果。“在所有使用场景中——包括直接案件告破、调查线索、跨司法管辖区协助以及现场部署——Flock技术全年支持了约100万起调查和事件，”该公司在博客文章中说道。
犯罪数据专家表示，该公司关于破案的统计声明并不可靠。
“这东西的可信度和印它的纸差不多，”曼哈顿研究所高级研究员查尔斯·费恩·莱曼说道，他本人是该技术的支持者。“……但我认为如果你深入研究学术研究，仍能发现其确实有效果——只是效果没那么显著。”
Flock在其2025年影响力普查博客文章中的脚注和警告称，这些数据“属于方向性估算，而非经审计的总计”，其方法学“尚未成型”。
该公司的说法已经出现在地方政府文件和审议中，各市政当局正在权衡这款监控技术的成本与收益。
今年夏天，加利福尼亚州科斯塔梅萨市议会就是否继续与Flock签订每年151800美元的合同进行审议时，市记录显示，其警察部门出具了正面推荐意见，其中部分依据是“Flock技术协助破获了2025年约20%的已告破案件”这一说法。
7月21日，科斯塔梅萨市议会以6票赞成、1票反对的结果保留了摄像头，但与Flock重新协商了合同条款。
即便其首席执行官按照网站上的所有附带警告来描述Flock的统计说法，犯罪学家也表示，其方法学仍存在更大的问题。
犯罪学家、Gainwell Technologies的数据科学家安德鲁·惠勒曾发表过关于自动车牌识别技术的研究，他表示，Flock的研究存在一个主要问题：即便Flock被用于协助某起案件，“这也未必意味着，如果没有Flock摄像头，这起案件就无法告破”。
他表示，Flock对2025年犯罪情况的研究并未使用既定的实证方法来考量这种反事实情况。
“基于这项特定研究，‘100万起案件告破’的说法很可能是严重高估，”惠勒说道。
存在争议的统计声明
这并非Flock的统计数据首次受到审查。
2024年，该公司开始宣称其技术“有助于破获美国10%的报案案件”（这与其后来的说法不同，后者称2025年客户辖区内约20%的已告破案件得到了Flock技术的协助）。
这一10%的说法在2024年2月的一份新闻稿中被公开推广，该新闻稿与两名Flock员工撰写的白皮书发布相关。
但实际上，这份由Flock撰写的白皮书文本中并未包含10%的说法。
相反，该白皮书得出的结论是，每增加一名宣誓警官配备的摄像头数量，与该机构车牌阅读器协助的案件告破率增加9.1个百分点相关。
白皮书称，其“经两名学者咨询后完成”——德克萨斯基督教大学的约翰尼·南和德克萨斯大学泰勒分校的理查德·赫尔弗斯。
该白皮书基于Flock从477家机构收到的调查反馈进行分析，其中123家机构提供了用于最终模型的数据。由于这些机构并非随机选取，无法确定它们能代表美国18000家执法机构。
惠勒当时在一篇博客文章中批评了该白皮书的方法学，并称Flock协助破获全国10%犯罪案件的结论“荒谬可笑”。
作为该白皮书的两名顾问之一，南在2024年告诉404 Media，他和赫尔弗斯协助完成了统计计算和格式排版，但关于结论，他表示“我个人会用截然不同的方式处理”。
在回应哥伦比亚广播公司新闻的提问时，南同样收回了对2024年白皮书结论的支持。
“我确实看过该论文的一个版本，也稍微讨论过，但我们并不了解其方法和数据来源，所以没有深入研究，”他说道。“我个人不想深入探讨其有效性或对犯罪率的影响，这对任何犯罪学家来说都是一个非常复杂的话题。”
相比之下，Flock在2024年的发布中引用了南的一句话，称赞该公司“在研究过程的每个阶段都完全透明、易于沟通”。
今年8月，在哥伦比亚广播公司新闻的采访中被问及Flock的全国破案说法是否有同行评审研究支持时，兰利表示同行评审研究并非该公司说法的一部分。
“是的，没有同行评审研究，我们也没有声称它经过了同行评审，”兰利说道。
他补充称，上个月的数据有“一个无可辩驳的数据点”。
“就在上个月的7月，我们检测到1025名失踪人员关联车辆，22000辆被盗车辆，”他说道。“这些检测会带来调查结果。”
但检测与失踪人员相关的车辆，与找到失踪人员本身是两回事。Flock曾表示，其车牌摄像头不使用面部识别技术。
当被问及支持这项“无可辩驳”的7月数据的原始信息时，卢贝尔告诉哥伦比亚广播公司新闻，这两项统计数据“基于与美国国家犯罪信息中心（NCIC）数据库和美国国家失踪与受虐儿童中心的安珀警报系统数据库记录相关的独特车辆检测”。
内部矛盾
截至8月下旬，Flock官网发布的一系列破案统计数据难以相互印证。在哥伦比亚广播公司新闻询问后，其中几项数据已被移除。
“Flock技术目前每天协助破获约2000多起案件，占美国所有报案犯罪的10%以上，”该公司在2026年公司概况介绍的第三页中写道，直到不久前，该介绍还是其官网“媒体资料包”的一部分。
同一份概况介绍的另一页则给出了完全不同的数字：“借助Flock技术，每周可破获2800多起案件”（这相当于每天约400起案件）。
在其他地方，该公司的常见问题解答页面同时声称“平均每天借助Flock技术破获1000多起案件”和“平均每周借助Flock技术破获2200多起案件”（后一数字相当于每天约314起案件）。
“你所询问的统计数据来自不同时间发布的材料，反映了不同的时间段、数据集和方法学，”卢贝尔在书面回复中回应这些不一致之处时说道。
本月早些时候哥伦比亚广播公司新闻询问后，Flock的2026年公司概况介绍以及此前存放在其官网“媒体资料包”链接的公开谷歌云端硬盘文件夹中的其他文件已无法访问。
该公司的常见问题解答页面也进行了更新，移除了“每天破获1000多起案件”的说法。
“这是我可以做出的权衡”
Flock的破案说法与其快速增长和融资的时期相吻合。该公司已从硅谷筹集了超过10亿美元资金，据《Inc.》杂志的一篇文章报道，其估值约为84亿美元。
在不同的公开声明中，兰利对这一百万起案件的统计数据的表述有所不同。有时，他完全没有任何背景地直接引用该数据。
“我们从未想过，就像我们产生的巨大影响力一样，对吧？”兰利在最近接受哥伦比亚广播公司新闻采访时说道。“仅去年一年，就有超过100万起刑事案件告破。”
在2026年4月的一场TED演讲中，主持人询问兰利Flock技术被滥用的案例，该演讲目前已获得超过500万次观看。
兰利承认存在隐私滥用的情况，随后转而谈到Flock的益处。
“我看到这100多万起案件，就会想，好吧，这是我可以做出的权衡，”兰利在回答时未对这一百万起案件的说法做出任何进一步解释。“但这是我的选择。”
话音刚落，TED演讲主持人向兰利表示感谢，当首席执行官走下台时，观众鼓掌致意。
Flock’s CEO claims its tech “solved” 1 million crimes last year. The company’s own methodology undercuts that claim
2026-09-30T14:07:00-0400 / CBS News
Faced with recent questions about the privacy costs of its technology, the surveillance camera startup Flock Safety and its CEO have deployed a tidy statistic to tout its benefits.
“Last year we helped clear just over a million crimes in America,” Flock CEO Garrett Langley said in a March 2026 podcast interview.
“Over a million cases were solved last year with this technology,” Langley told the auto industry publication The Drive, “and the level of abuse is incredibly low relative to that.”
“Today, Flock technology assists your officers in more than one million crimes solved annually,” Langley said in a blog post last month announcing new privacy updates.
But a CBS News review finds the claim that Flock helped solve a million crimes in 2025 is not supported by the available evidence.
In a statement for this report, Flock spokesman Paris Lewbel cast the statistic in a different framing. In addition to case clearances, he said, it includes “investigative leads, cross-jurisdictional assistance, searches conducted during active investigations, and other Flock deployments.”
“The most precise way to describe our 2025 national estimate is that Flock technology supported approximately 1 million investigations and incidents,” Lewbel said.
The company did not answer whether Langley had been imprecise in claiming Flock solved one million crimes in 2025, and did not provide CBS News with underlying data supporting that claim.
A Flock camera in Norwood, Massachusetts. CBS Boston
Flock first explained the source of its claims of one million crimes solved in a May 2026 blog post announcing the results of its 2025 Impact Census, in which about 700 law enforcement agencies that use Flock’s system participated. Flock says agencies that formally tracked outcomes supplied raw counts, while others provided percentage estimates. The blog post also said the study found 20% of cleared cases in customer jurisdictions in 2025 were assisted by Flock technology.
In explaining the 20% finding, the company said 307 agencies supplied percentage estimates. Those responses were then referenced with FBI crime and clearance data for each responding agency.
The “one million crimes” statistic was a separate and broader calculation. “Across all use cases — direct case clearances, investigative leads, cross-jurisdictional assists, and field deployments — Flock supported approximately 1 million investigations and incidents during the year,” the company said in the blog post.
Experts in crime data say the company’s crime-solving statistical claims are not reliable.
“It’s worth about the paper it’s printed on,” said Charles Fain Lehman, a senior fellow at the Manhattan Institute who is a proponent of the technology. “…But I think if you look under the hood of the academic research, that still points you in a direction of there’s an effect — it’s just a smaller effect.”
Footnotes and caveats on Flock’s blog post regarding its 2025 Impact Census say these figures are “directional estimates rather than audited totals” and that its methodology is “not a finished product.”
The company’s claims have surfaced in local government documents and deliberations as municipalities grapple with the costs and benefits of the surveillance technology.
When the city council in Costa Mesa, California, considered whether to continue its contract with Flock this summer at an annual cost of $151,800, city records show, its police department issued a positive recommendation that cited in part the claim that “Flock technology assisted in solving an estimated 20% of cleared cases” in 2025.
On July 21, Costa Mesa City Council voted 6-1 to keep the cameras but renegotiate the city’s contract with Flock.
Even if its CEO described Flock’s statistical claims with all the caveats included on its website, criminologists say there are bigger issues with its methodology.
A major issue with Flock’s study is that even if Flock is used to help with a case, “it doesn’t necessarily mean in a counterfactual world where Flock cameras didn’t exist, the case wouldn’t have been solved,” said Andrew Wheeler, a criminologist and data scientist at Gainwell Technologies who has published research on automated license plate technology.
Flock’s study of 2025 crime, he said, did not use established empirical methods to account for this counterfactual.
The “a million crimes solved” is probably a gross over-estimate based on that particular work,” Wheeler said.
Disputed statistical claims
It is not the first time Flock’s statistics have come under scrutiny.
In 2024, the company began claiming that its technology “is instrumental in solving 10% of reported crimes in the U.S.” (This is distinct from its later claim that an estimated 20% of cleared cases in 2025 were assisted by Flock technology in customer jurisdictions.)
The 10% claim was promoted publicly in a February 2024 press release connected to the release of a white paper authored by two Flock employees.
However, that 10% claim is not actually contained within the text of the Flock-authored white paper.
Rather, the paper concludes that every additional camera per sworn officer is associated with a 9.1-percentage-point increase in the agency’s license plate reader-assisted clearance rate.
The white paper says it was “created with the consultation” of two academics — Johnny Nhan of Texas Christian University and Richard Helfers of the University of Texas at Tyler.
The white paper is an analysis based on survey responses Flock received from 477 agencies, 123 of which supplied data used in the final model. Because the agencies were not randomly selected, it was not established that they were representative of the 18,000 law enforcement agencies in the United States.
Wheeler, in a blog post at the time, criticized the white paper’s methodology and called the conclusion that Flock helped solve 10% of crimes nationwide “ridiculous.”
Nhan, one of the two consultants for the white paper, told 404 Media in 2024 that he and Helfers helped with statistical calculations and formatting, but regarding the conclusions, he said he “personally would have done things much differently.”
In response to questions from CBS News, Nhan similarly backed away from the 2024 paper’s conclusions.
“I did get to look at a version of the paper and did discuss a little bit, but the methods and data sources were not known to us so we didn’t dive into it,” he said. “I personally didn’t want to delve into its effectiveness or impact on crime rates, which is a very complex topic for any criminologist.”
By contrast, Flock’s 2024 release attributed a quotation to Nhan praising the company as “fully transparent and accessible during every stage of the research process.”
Garrett Langley, founder and CEO of Flock Safety Aaron Schwartz/Bloomberg via Getty Images
Asked in a CBS News interview in August if Flock’s national crime-solving claims are supported by peer-reviewed studies, Langley said peer-reviewed research was not part of the company’s claims.
“No, there’s no peer-reviewed study, and we’re not articulating that it was peer-reviewed,” Langley said.
He added that there was “one indisputable data point” from the prior month’s data.
“In July of this year, just last month, 1,025 missing people detected, 22,000 stolen cars detected,” he said. “Those detections lead to results.”
The detection of a vehicle associated with a missing person case, however, is different from the detection of a missing person. Flock has said its license plate cameras do not use facial recognition technology.
Asked for underlying data supporting the “indisputable” July figures, Lewbel told CBS News those two statistics “are based on unique vehicle detections involving vehicles associated with records in the National Crime Information Center (NCIC) database and the National Center for Missing & Exploited Children’s AMBER Alert database.”
Internal inconsistencies
As of late August, Flock’s website featured a series of crime-solving statistics that could not easily be reconciled. Several of those statistics were removed following an inquiry from CBS News.
“Flock technology is currently being used to solve approximately 2,000+ crimes per day, accounting for over 10% of all reported crime in the United States,” the company said on the third page of a 2026 Company Fact Sheet that was, until recently, part of a media kit available on its website.
A separate page of the same fact sheet provided an entirely different figure: “2,800+ crimes solved weekly with the help of Flock Technology.” (That’s about 400 crimes per day.)
Elsewhere, the company’s FAQ page stated both “On average, 1,000+ crimes solved per day with the help of Flock Technology” and “On average, 2,200+ crimes solved per week with the help of Flock technology.” (The latter figure is about 314 crimes per day.)
“The statistics cited in your questions come from materials published at different points in time and reflect different time periods, datasets, and methodologies,” Lewbel said in written response to questions about the inconsistencies.
Following questions from CBS News earlier this month, Flock’s 2026 Company Fact Sheet and other documents which had been stored in a publicly-viewable Google Drive folder linked from the “Media Kit” section of the company’s website became unavailable.
The company’s FAQ page also changed, removing the claim of “1,000+ crimes solved per day.”
“A trade-off that I can make”
Flock’s crime-solving claims have coincided with a period of rapid growth and investment. The company has raised more than $1 billion from Silicon Valley and, according to an article in Inc., is valued at approximately $8.4 billion.
In different public statements, Langley has altered the framing around the million-crimes statistic. At times, he has cited it without any context at all.
“We never would have guessed, just like the sheer impact we’ve had, right?” Langley said in his recent interview with CBS News. “So last year alone, over a million criminal cases.”
In an April 2026 TED Talk that has now been viewed more than 5 million times, a host asked Langley about instances in which Flock’s technology has been abused.
Langley conceded there have been instances of privacy abuse before pivoting to Flock’s benefits.
“I look at the million-plus cases and go, OK, this is a trade-off that I can make,” Langley said in reply, without providing any further description of the one-million figure. “But that’s my choice.”
With that, the TED Talk host thanked Langley and the audience applauded as the CEO walked off stage.
发表回复