Flock says its new tool will help identify police abuse, but hasn’t explained how it works
Flock says its new tool will help identify police abuse, but hasn’t explained how it works
Flock 称其新工具将有助于识别警察滥用职权行为，但尚未解释其工作原理
Surveillance technology startup Flock announced a series of new policies and tools on Thursday, claiming they will help curb abuse of its automated license plate camera reader systems and help keep its law enforcement customers accountable. 监控技术初创公司 Flock 周四宣布了一系列新政策和工具，声称这些举措将有助于遏制对其自动车牌识别摄像系统的滥用，并帮助其执法部门客户承担责任。
Flock is now recommending police departments retain data for seven days rather than the previous 30 days. Along with this change, the company introduced what it calls “Evidence Mode,” which can be turned on when customers need to retain data for longer periods in “exceptional cases” and requires a specific case number. Customers will also now be able to set limits on the data they share with other customers based on the type of offense that’s under investigated. Flock 目前建议警察部门将数据保留时间从之前的 30 天缩短至 7 天。伴随这一变化，该公司还引入了所谓的“证据模式”（Evidence Mode），当客户在“特殊情况”下需要更长时间保留数据时，可以开启该模式，并要求提供具体的案件编号。此外，客户现在还可以根据正在调查的犯罪类型，对其与其他客户共享的数据设置限制。
The company also highlighted a tool called “Audit Assistance” as one of the key changes. Flock launched the tool in April. Since then, Flock said “more than one-third” of its customers have turned it on, and the startup is now requiring all customers to enable it by the end of the year. Flock said Audit Assistance “detects abnormal activity and flags it for Administrator review,” according to a blog posted released Thursday. If that happens, the tool can automatically lock out users who have been flagged until an administrator intervenes. 该公司还将一款名为“审计辅助”（Audit Assistance）的工具作为关键变革之一进行了重点介绍。Flock 于今年 4 月推出了该工具。Flock 表示，自那时起，已有“超过三分之一”的客户启用了该功能，目前该公司要求所有客户在年底前必须启用它。根据周四发布的一篇博客文章，Flock 表示“审计辅助”功能可以“检测异常活动并将其标记以供管理员审查”。如果发生这种情况，该工具可以自动锁定被标记的用户，直到管理员介入处理。
According to Flock, the tool has already helped customers spot and stop abuse. Several executives have touted the capabilities of Audit Assistance. But despite its stated importance, Flock has not explained in detail how Audit Assistance actually works. Some official comments provide some understanding of what Audit Assistance can do; it gets vague when trying to determine how it does it. 据 Flock 称，该工具已经帮助客户发现并制止了滥用行为。几位高管都吹捧了“审计辅助”的功能。但尽管其重要性不言而喻，Flock 尚未详细解释“审计辅助”的具体工作原理。一些官方评论提供了一些关于该工具功能的了解，但在试图确定其运作方式时，表述却变得模糊不清。
For instance, Flock’s head of trust and compliance Ashley Haber said in a video that the Audit Assistance can flag “what may look like an odd search history from a specific user,” helping customers understand how their users are operating the system. In another video, Flock co-founder Paige Todd said “Audit Assistance actually flags unusual patterns early. Think a user searching the same plate repetitively for more than 30 days.” And Major Patrick Krieg of the Dunwoody, Georgia Police Department, a Flock customer, described the feature in a press release as “an algorithm that has notified us of any type of bias.” 例如，Flock 的信任与合规主管 Ashley Haber 在一段视频中表示，“审计辅助”可以标记“特定用户看似奇怪的搜索历史”，帮助客户了解其用户如何操作该系统。在另一段视频中，Flock 联合创始人 Paige Todd 表示：“‘审计辅助’实际上能及早标记异常模式。比如用户在 30 多天内重复搜索同一个车牌。”而作为 Flock 客户的佐治亚州邓伍迪警察局少校 Patrick Krieg 在一份新闻稿中将该功能描述为“一种能通知我们任何类型偏见的算法”。
Flock describes Audit Assistance as a feature that can “surface atypical activity early and review it quickly with a clear, documented workflow.” What’s unclear — and TechCrunch has asked for answers about — is how this “algorithm” actually works, what data is it trained on, and if it was developed using an large language model. TechCrunch also asked what kind of patterns is it designed to surface and what exactly gets flagged as “abnormal activity”? Flock 将“审计辅助”描述为一种能够“及早发现非典型活动，并通过清晰、有据可查的工作流程进行快速审查”的功能。目前尚不清楚——TechCrunch 也已就此寻求答案——的是，这个“算法”究竟是如何工作的，它基于什么数据进行训练，以及它是否使用大语言模型开发。TechCrunch 还询问了它旨在发现什么样的模式，以及究竟什么会被标记为“异常活动”？
According to Flock, Audit Assistance is not based on [Machine Learning] or AI. The company told TechCrunch that “It is a data tool that flags atypical search patterns for further review. One example that would trigger a flag is if a user searches for the same plate using multiple different case codes, as this could potentially indicate abuse. A flag does not mean that it is definitely abuse — it only means it warrants further investigation.” 据 Flock 称，“审计辅助”并非基于机器学习或人工智能。该公司告诉 TechCrunch：“它是一个数据工具，用于标记非典型搜索模式以供进一步审查。触发标记的一个例子是，如果用户使用多个不同的案件代码搜索同一个车牌，因为这可能表明存在滥用行为。被标记并不意味着一定是滥用——它只意味着该行为值得进一步调查。”
Numerous questions remain unanswered, however, including whether Flock has any statistics or data that shows how effective the tool is, what its false positive and false negative rates are, and if it has been audited by a third party? 然而，仍有许多问题悬而未决，包括 Flock 是否有任何统计数据或资料来证明该工具的有效性，其误报率和漏报率是多少，以及它是否经过了第三方审计？
Privacy experts and Flock critics are not impressed by Audit Assistance. The American Civil Liberties Union senior policy counsel Chad Marlow wrote in response to Flock’s announcement that “there is no evidence that the tool works consistently” to address abuse. “While dozens of officers have recently been arrested, fired, or otherwise disciplined for misusing Flock for personal reasons, Flock claims these arrests are proof its auditing tool works,” Marlow wrote. “However, unless we know the number of officers misusing the system, we cannot conclude if Flock and its auditing tools are catching 95% of violators or 5%. Flock needs to have its auditing tool analyzed by an independent evaluator to determine its actual effectiveness. Until then, we don’t know if the tool is a real security measure or just window dressing.” 隐私专家和 Flock 的批评者对“审计辅助”并不买账。美国公民自由联盟（ACLU）高级政策顾问 Chad Marlow 在回应 Flock 的声明时写道，“没有证据表明该工具在解决滥用问题上能持续发挥作用”。Marlow 写道：“虽然最近有数十名警官因出于个人原因滥用 Flock 而被捕、解雇或受到纪律处分，但 Flock 声称这些逮捕证明了其审计工具是有效的。然而，除非我们知道滥用系统的警官总数，否则我们无法断定 Flock 及其审计工具是抓住了 95% 的违规者还是仅 5%。Flock 需要让独立评估机构对其审计工具进行分析，以确定其真实有效性。在此之前，我们不知道该工具是真正的安全措施，还是仅仅是装点门面。”
Cooper Quintin, a security researcher and senior public interest technologist at the Electronic Frontier Foundation, told TechCrunch that it may not even matter whether Audit Assistance works. “If it does work, law enforcement will figure out ways to get around it,” he said, adding that what matters in the end is whether abusers are caught and held accountable. “If there are no consequences for abuse and whoever [the tool] reports to is the same police agency that is doing the abuse, it’s just a fig leaf for Flock. The best way to ensure accountability is to pass laws constraining the use of this technology and requiring a warrant for its use.” 电子前沿基金会（EFF）的安全研究员兼高级公共利益技术专家 Cooper Quintin 对 TechCrunch 表示，“审计辅助”是否有效可能根本不重要。“如果它真的有效，执法部门也会想出绕过它的办法，”他说，并补充道，最终重要的是滥用者是否被抓获并追究责任。“如果滥用行为没有后果，而且（该工具）报告的对象正是实施滥用的同一个警察机构，那么这只是 Flock 的一块遮羞布。确保问责制的最佳方式是制定法律，限制这种技术的使用，并要求在使用时必须获得搜查令。”
Rachel Levinson-Waldman, the director of the Brennan Center’s Liberty and National Security Program, told TechCrunch that making Audit Assistance mandatory is a step in the right direction, and that “in theory” the tool is “a great idea.” But she also said the tool itself “needs to be audited” because there’s not enough information about how it works in practice. 布伦南中心（Brennan Center）自由与国家安全项目主任 Rachel Levinson-Waldman 告诉 TechCrunch，将“审计辅助”设为强制性功能是朝着正确方向迈出的一步，并且“从理论上讲”，该工具是一个“好主意”。但她同时也表示，该工具本身“需要接受审计”，因为目前关于它在实践中如何运作的信息还不够充分。
Flock’s announcements come after years of reported incidents of its customers repeatedly abusing its technology. Earlier this week, three former Georgia sheriff’s deputies were arrested and accused of using Flock cameras to stalk people with whom they had a “personal relationship at the time,” as the Bibb County Sheriff’s Office said in a press release. In this case, the Bibb County sheriff David Davis specifically called out Flock’s new tool Audit Assistance for uncovering the misuse. Flock 的这些公告是在其客户多年来反复滥用其技术的报道之后发布的。本周早些时候，据比布县（Bibb County）治安官办公室在一份新闻稿中称，三名前佐治亚州治安官副手被捕，并被指控利用 Flock 摄像头跟踪他们“当时有私人关系”的人。在此案中，比布县治安官 David Davis 特别提到了 Flock 的新工具“审计辅助”在揭露此次滥用行为中的作用。