来源：《华盛顿邮报》
原文刊登日期：2021年6月3日
Facial recognition software is full of peril, but it’s also full of potential. With the powerful technology still in its infancy and regulation next to nonexistent, now is the time to write rules that would nudge its development in a useful rather than a dystopian direction.
人脸识别软件充满了危险，但也充满了潜力。鉴于这项强大的技术仍处于起步阶段，监管几乎不存在，现在是时候制定规则，推动其向有益而非反面乌托邦的方向发展。
The Metropolitan Washington Council of Governments announced last month that its National Capital Region Facial Recognition Investigative Leads System, which attempts to match suspects’ faces to those in a database of mugshots, would cease to operate, in part because a new Virginia law requires approval from the state legislature for any local law enforcement system. Around the same time, Amazon extended indefinitely the global ban the company instituted last summer on police use of its software. There’s an obvious cause for caution: Police have wrongfully arrested at least three men based on bad matches, and tales of gross misuse among law enforcement abound.
华盛顿大都会政府委员会上个月宣布，其国家首都地区人脸识别调查线索系统将停止运行，该系统试图将嫌疑人的面部在人脸数据库中进行特征匹配，停运的部分原因是弗吉尼亚州的一项新法律要求任何地方执法部门都必须得到州立法机关的批准。大约在同一时间，亚马逊无限期延长了该公司去年夏天对警察使用其软件的全球禁令。这样警惕有明显的理由：警方至少因匹配错误逮捕了三名男子，执法人员严重滥用权力的传闻比比皆是。
Congress should impose a temporary stopping on government use of facial recognition software until it has time to erect guardrails that would prevent these abuses. Those guardrails should include setting minimum performance standards that require high confidence thresholds for matches and restrict real-time surveillance to public emergencies. But as long as Congress dithers, states, cities and companies are right to fill the rules gap, shielding civilians from overreach. Unchecked use of the technology threatens not only individuals whom the technology wrongly matches, but also society in general. Critics rightly worry about national capital region officials deploying facial recognition software to hound participants in last summer’s largely peaceful protests at Lafayette Square.
国会应该暂时停止政府使用人脸识别软件，直到有时间制定措施防止滥用。这些防护措施应包括设定最低性能标准，要求对匹配有很高的置信阈值，并限制对公共紧急情况的实时监控。但只要国会还在犹豫，各州、城市和企业就有权利填补规则的空白，保护平民免受越权的侵害。对技术的无限制使用不仅威胁到被技术错误匹配的个人，而且威胁到整个社会。批评人士的担忧不无道理，去年夏天，首都地区官员部署人脸识别软件来追踪拉法叶广场大体和平的抗议活动参与者。
Yet, rule-makers must also avoid overreach as they respond to these threats. The technology should be regulated, not shut down. Law enforcement officers have harnessed facial recognition in bank robbery and human trafficking cases. A good recognition algorithm, appropriately trained and reviewed, could do a better job correctly identifying a suspect — and indeed, exhibit less prejudice — than an officer flipping through mugshots.
然而，在应对这些威胁时，规则制定者也必须避免反应过度。这项技术应该受到监管，而不是封杀。执法人员已经在银行抢劫和人口贩卖案件中使用了人脸识别技术。一个良好的识别算法，经过适当的训练和审查，可以比警察翻看面部照片更好地正确识别嫌疑人——并且，表现出更少的偏见。
Congress’s inaction so far leaves Americans unprotected. Authorities will continue to encroach on civil liberties, and at the same time, those excesses threaten to create a backlash that clamps down too hard on a tool that could aid important investigations and keep people safer. Before that happens, legislators must ensure that law enforcement treats powerful software with care and caution.
到目前为止，国会的不作为让美国人得不到保护。行政当局将继续侵犯公民自由，与此同时，政府的越轨行为可能会引发反弹，对一种可能有助于重要调查和确保人民安全的工具进行过度打压。在这种情形发生之前，立法者必须确保执法部门小心谨慎地对待这种功能强大的软件。