汽车自动驾驶技术研发、测试已多年，但在2024年进入了公众视野中的“发展元年”。7月，百度旗下的“萝卜快跑”无人车在武汉等地开启大规模投放，引发舆论探讨；10月，特斯拉首发无人驾驶出租车Cybercab，自动驾驶延展为全球范围内的顶流话题。

• 2024年自动驾驶产业加速融合日常生活，迎来IPO热潮
• 产业仍面临准入、数据、事故归责等多重法律挑战

截止今年8月底，中国已累计发放自动驾驶汽车测试号牌1.6万张，开放公共测试道路3.2万公里，自动驾驶正加速走入日常生活。

与此同时，伴随中国自动驾驶相关公司商业化压力加增，2024年该领域的融资活动也十分频繁。继禾赛科技、知行科技登陆资本市场后，今年8月，“自动驾驶芯片第一股”黑芝麻完成香港上市；10月，智能驾驶芯片公司地平线完成年内港股最大科技IPO，文远知行则登陆纳斯达克，成为“全球通用自动驾驶第一股”。

目前，该领域的小马智行、纵目科技和Momenta等领军公司也已提交了上市申请。

此外，大型车企也在加入自动驾驶版图。据咏竹坊报道，比亚迪、小鹏已在自动系统内部开发上投入大量资金；大众也与地平线设立合资公司，并向其投资了24亿欧元。

技术演进

世辉律师事务所合伙人王新锐律师的客户集中于科技领域，他也一直关注并服务于中国自动驾驶产业的发展。王律师总结道：“目前自动驾驶产业链可分为互联网大厂、造车新势力和自动驾驶创业公司几类，传统整车厂也在奋起直追。”

“我们的观察是：每个类别都在分化，并出现明显的领跑者。从融资和商业化角度，百花齐放的阶段已经结束，接下来会进入比较残酷的淘汰赛。”他说，“其中的一大变数是由特斯拉引领的端对端（end-to-end）自动驾驶技术，国内公司也都在仿效、追赶这一技术路线，可能会带来行业洗牌。”

王律师解释道，此前境内外主流厂商使用的主要是“基于规则”（Rule-Based）的自动驾驶方案。根据百度智能驾驶事业部曾做出的分享，rule-based需要先通过车辆传感器获取动态信息，再加入高精地图等静态信息，共同构建出对外部世界的描述，基于此做出决策，再对车辆下达控制命令，其所涉及的系统复杂，成本和算力要求也非常高。

2016年前后，新的端对端方案开始出现，其更接近于人的驾驶习惯——即看到人类驾驶员所看到的场景，继而经过神经网络处理后产生类似于人类的驾驶行为。

“特斯拉的完全自动驾驶（full self-driving, FSD）是端对端方案的代表。”王律师说，“端到端方案由模型基于海量的真实驾驶数据自动学习而来，将直接根据实际情境做出驾驶应对，更加趋近于生成式人工智能。”

不过，“端对端方案会带来更多的黑箱问题及可追责性方面的挑战，这些都有待未来法律法规的持续发展与最新的技术和风险相适配”，王律师坦言。

法律问题与监管现状

自动驾驶的普及无疑将极大提升出行便利，但其关乎公共交通，牵涉人身安全与公众福祉，其发展因此也呼求监管的快速跟进。

“目前中国围绕自动驾驶的产业发展、道路测试、交通安全、网络和汽车数据、产品准入、高精地图等均已开展立法工作，且已有大量文件出台。”王新锐律师说。不过自动驾驶行业在几个维度依然面临着比较突出的法律问题。

首先是行业准入。作为新质生产力的代表产业，中国正积极推动自动驾驶行业发展并颁发相关牌照，但“获取自动驾驶道路测试牌照和无人驾驶汽车市场准入许可仍需要满足一系列技术标准和安全要求，对企业来说是一个重要的法律挑战”。

其次是一旦出现安全事故的法律责任主体问题。王律师说：“根据自动驾驶汽车的自动化等级——如L4、L5，不同的责任主体可能包括制造商、软件开发者、车辆所有者等，当前法律体系难以清晰划分各方责任，导致在事故发生时责任归属模糊。”

2021年公安部起草公布《道路交通安全法（修订建议稿）》时，“第155条曾首次对自动驾驶汽车的道路测试通行、事故责任分担等内容做出规定。不过，最终这条‘自动驾驶条款’并未出现在正式发布的修订案中”。

相较而言，地方层面的立法更加积极，并且呈现出先行先试的态度。“目前全国已有超过50个城市出台自动驾驶相关的地方政策，其中深圳、上海、江苏、杭州等多省市已制定自动驾驶相关地方立法，并对一些亟需解决的问题做出了回应。”例如《深圳经济特区智能网联汽车管理条例》《北京市自动驾驶汽车条例（征求意见稿）》等，都对自动驾驶汽车在汽车交通事故中侵权责任分配问题进行了明确。

不过整体而言，目前中国相关立法的概况是，“中央层面立法仍以发布引导政策和具体应用领域的规范性文件为主，基础性规则文件较为缺乏。鉴于上位法尚未修改，地方立法仅能在有限空间内做出少量调整，较难进行突破性规定”，王律师指出。

自动驾驶行业面临的第三重法律挑战则在于，由于需要由车辆和系统进行自主判断，因此“在运行过程中会收集大量用户和环境数据，会引发个人隐私和数据安全的问题”。

就此，王律师说，为自动驾驶企业提供全生命周期数据合规服务目前是外部律师的重要工作内容之一。“包括座舱数据、测绘数据、个人信息等数据合规处理；面对运营过程中的生产、经销、售后维保等全场景，协同多部门以及多国公司开展个人信息处理情况评估；建立企业数据合规内部管理制度；车联网服务车机端及App端应用功能及相关数据问题，以及数据相关协议文本修改起草；和企业出海时的境外数据合规问题等。”

数据之外，目前外部律师提供的其他服务也具备一定独特性。例如人工智能合规，包括“人工智能相关算法技术安全、算法备案、内容治理风控、大模型合规框架搭建等问题”，以及技术转让协议相关的知识产权问题、导航电子地图制作相关的测绘资质等车企常见业务场景服务。

下一步，在自动驾驶技术飞速革新的背景下，“我们理解，法规与技术的协同发展将是值得关注的主要问题”，王律师预期道。

Autonomous Driving Hits the Fast Lane Amid Tech Advances and Legal Challenges

• Autonomous driving integrates into daily life and sees a surge in IPOs. 
• Legal challenges remain in licensing, data use, and liability for accidents. 

The year 2024 has been a watershed for autonomous vehicles. In July, Baidu’s Apollo Go began large-scale deployment of driverless cars in cities like Wuhan, sparking widespread public debate. By October, Tesla launched its Cybercab, a fully autonomous taxi, thrusting self-driving technology into the global spotlight.

By the end of August, China had issued 16,000 autonomous vehicle test licenses and opened 32,000 kilometers of public testing roads, signaling rapid integration of autonomous driving into everyday life.

Meanwhile, as commercialization ramps up, China's autonomous driving sector has seen a flurry of financing activity in 2024. Following the stock market debuts of Hesai Technology and Zhixing Technology, August saw Black Sesame Technologies—the "first autonomous driving chip stock"— listed in Hong Kong. In October, Horizon Robotics completed the largest tech IPO of the year on the Hong Kong Stock Exchange, while WeRide became the "first global general autonomous driving stock" with its Nasdaq listing.

Leading players Pony.ai, Zongmu Technology, and Momenta have also been listed or filed for IPOs, and major automakers are joining the fray. BYD and Xpeng have invested heavily in in-house autonomous systems, while Volkswagen has formed a joint venture with Horizon Robotics, committing €2.4 billion to the collaboration.

Technological Evolution

According to Raymond Wang, a partner at Shihui Partners, the autonomous driving industry is divided among internet giants, emerging automakers, and autonomous driving startups, with traditional automakers playing catch-up.

“Each segment is fragmenting, with clear frontrunners emerging. A fierce elimination race is ahead,” says Wang. He notes that Tesla’s pioneering end-to-end autonomous driving technology is reshaping the industry, prompting domestic companies to follow suit and potentially triggering a shakeout.

Wang explains that traditional rule-based solutions used by major autonomous driving companies rely on vehicle sensors and high-definition maps to create a static and dynamic representation of the external environment, which guides decision-making and vehicle control. However, these systems are complex, costly, and computationally intensive.

In contrast, end-to-end solutions—introduced around 2016—replicate human driving behavior. These systems process real-world driving scenarios through neural networks, learning from vast amounts of real driving data to generate responses akin to human drivers.

“End-to-end systems operate like generative AI, responding directly to real-world contexts. However, they also pose challenges, such as ‘black box’ issues and difficulties in assigning liability, which will require legal frameworks to adapt alongside technological advancements,” Wang adds.

Legal and Regulatory Hurdles

The promise of autonomous driving is as transformative as it is complex. Yet, its journey to widespread adoption comes with significant regulatory and legal challenges, particularly in ensuring public safety and addressing accountability.

China has made significant progress in legislating for autonomous driving, covering areas such as industry development, road testing, traffic safety, data security, product approvals, and high-definition mapping. “A substantial number of regulatory documents have already been issued,” sats Wang, however, he notes that critical legal challenges persist in several areas.

Industry access remains one of the hurdles. As a representative of emerging industries, autonomous driving has seen a push for growth, yet, Wang explains, “Securing road test licenses and market access for driverless vehicles requires meeting a series of technical and safety standards, presenting a significant legal challenge for companies.”

Another key issue lies in determining liability when accidents occur. “Depending on the level of automation—such as L4 or L5—the responsible party could be the manufacturer, software developer, or vehicle owner,” says Wang. “The current legal framework struggles to clearly allocate responsibility, leading to ambiguity in assigning liability after accidents.” 

Local governments have taken some proactive steps, with cities like Shenzhen and Beijing issuing regulations to clarify fault in autonomous driving-related incidents. Still, Wang observes, “The lack of foundational regulations at the central level and the constraints of existing laws make groundbreaking changes through local legislation difficult.”

Privacy and data security present another pressing concern. Autonomous vehicles, which rely on vast amounts of data to make independent decisions, must collect user and environmental information during operation. “This raises significant risks for personal privacy and data protection,” Wang says, adding that advising companies on full-lifecycle data compliance has become a core focus for external lawyers in this field.

Beyond data, legal services for autonomous driving companies increasingly encompass artificial intelligence compliance, including algorithm safety, regulatory filings, content governance, and frameworks for large-model compliance. Other issues include intellectual property in technology transfer agreements and mapping qualifications for navigation systems.

As autonomous driving technologies evolve rapidly, Wang predicts that the alignment of regulations with innovation will remain a central issue. “The collaboration between law and technology will be critical to watch in the future,” he says.