目前尚无充分证据证明自动驾驶出租车在整体上比人类驾驶更安全。美国每年约有40,000人死于交通事故，而支持者声称Waymo等公司的自动驾驶技术已经降低风险，甚至“拯救生命”。然而，多名独立研究者指出，现有数据不足以得出结论。Waymo累计行驶约1.27亿英里，却已涉及2起致命事故；按“每英里死亡率”计算，这一比例高于美国人类驾驶平均约1.23亿英里对应1起死亡。统计学研究表明，需要数亿甚至数百亿英里的数据，才能可靠比较自动驾驶与人类驾驶的安全性。

数据本身也存在系统性偏差。Waymo车辆大多在限速35英里/小时以下的城市道路运行，且在加州约50%的行驶里程为空载状态，这使其与普通人类驾驶情境并不对等。此外，自动驾驶系统频繁软件更新，可能显著改变风险水平，却仍被合并进同一安全统计。远程人工接管员的介入程度缺乏透明披露，也可能低估事故风险。近期旧金山停电事件中，多辆Waymo车辆在路口停滞，凸显系统在非常规情境下的脆弱性。

即便假设自动驾驶单次行程更安全，也未必降低整体死亡人数。若自动驾驶使总行驶里程增加30%，而事故率仅下降20%，总体事故数仍会上升约4%。相比之下，许多更廉价、已被验证有效的措施——如限速技术、缩小大型车辆、建设受保护自行车道和人行道、扩大公共交通——能直接减少死亡且不会刺激额外出行。芬兰赫尔辛基和挪威奥斯陆已实现一年内零行人和骑行者死亡，且完全未依赖机器人出租车。

We still lack sufficient evidence to conclude that robotaxis are safer than human drivers overall. About 40,000 people die annually in US traffic crashes, and proponents argue that companies like Waymo are already reducing this toll. Yet multiple independent researchers say current data is inadequate. Waymo has logged roughly 127 million miles but has been involved in two fatal crashes; measured by deaths per mile, this rate exceeds the US human-driving average of about one fatality per 123 million miles. Statistical studies suggest hundreds of millions or even hundreds of billions of miles are needed for reliable safety comparisons.

The data is also structurally biased. Most Waymo driving occurs on urban streets with speed limits of 35 miles per hour or less, and in California about 50% of its mileage is without passengers, making comparisons with typical human driving uneven. Frequent software updates can materially change risk but are still pooled into a single safety record. The extent of remote human teleoperation remains opaque and may suppress reported risk. Recent San Francisco power outages, when multiple Waymo vehicles stalled at intersections, exposed weaknesses in unusual conditions.

Even if autonomous trips prove safer individually, total fatalities may not fall. If automation cuts crash risk by 20% but increases total driving by 30%, total crashes rise by about 4%. Meanwhile, cheaper, proven interventions—such as speed limiters, smaller vehicles, protected bike lanes, wider sidewalks, and expanded public transit—reduce deaths without inducing more driving. Cities like Helsinki and Oslo have recorded entire years with zero pedestrian or cyclist deaths, achieved through policy and infrastructure changes rather than robotaxis.