Autonomous Vehicle Pilots: Seoul’s Self-Driving Technology Testing and Regulatory Framework

South Korea has staked a national industrial claim on autonomous vehicle technology, and Seoul is the proving ground. The Ministry of Land, Infrastructure and Transport (MOLIT) has designated the capital as the primary deployment zone for Level 4 autonomous vehicle testing under the Autonomous Vehicle Act of 2020, and the Seoul Metropolitan Government has responded by constructing the most comprehensive urban AV testing ecosystem in East Asia. Sixty-seven autonomous vehicles hold active Seoul testing permits as of early 2026, operating across four designated test zones that span 128 km of public roads. Twelve companies — including Hyundai Motor, Samsung Electronics, Naver Labs, KT, 42dot (Hyundai subsidiary), Kakao Mobility, and four international entrants — maintain active test fleets. The accumulated test mileage exceeds 4.2 million km, with a disengagement rate that has improved from 14.7 per 1,000 km in 2022 to 2.1 per 1,000 km in 2025.

The question confronting Seoul is not whether autonomous vehicles will operate on its streets — they already do — but how the transition from controlled testing to commercial deployment will be managed in a dense, complex urban environment where 9.4 million people, 3.12 million registered vehicles, and 7,500 public buses share road space with increasingly capable machines.

The Regulatory Architecture: Korea’s Autonomous Vehicle Act

Korea’s regulatory framework for autonomous vehicles operates through the Autonomous Vehicle Act (enacted February 2020, amended January 2024), which establishes a five-level classification system aligned with SAE International standards:

Level 1-2: Driver-assistance systems requiring continuous human supervision. These are unregulated beyond standard vehicle-safety requirements and are present in approximately 34 percent of new vehicles sold in Korea. Level 3: Conditional automation where the vehicle handles all driving functions in specific conditions but requires a human driver ready to resume control. Hyundai’s Highway Driving Pilot system, available on the IONIQ 5 and Genesis GV80 since 2024, operates at Level 3 on Korean expressways. Level 4: High automation where the vehicle operates without human intervention in defined operational design domains (ODDs). This is the level at which Seoul’s test programmes operate. Level 5: Full automation in all conditions. No vehicle currently operates at Level 5 anywhere in the world.

The Act establishes three categories of operating permit. Temporary Test Permits allow AV testing on designated public roads with a safety driver present. Pilot Operation Permits allow limited commercial service (fare-charging passenger or freight transport) within geographically defined zones, with or without an onboard safety operator depending on the vehicle’s demonstrated safety record. Full Operation Permits — which no entity has yet received — will authorise unrestricted commercial AV operations on all public roads within the vehicle’s certified ODD.

The permit system is administered by MOLIT’s Autonomous Vehicle Division, with technical safety assessment delegated to the Korea Automobile Testing and Research Institute (KATRI) and the Korea Transportation Safety Authority (KOTSA). The safety assessment requirements for a Pilot Operation Permit include: minimum 100,000 km of incident-free test operation within the proposed service zone, disengagement rates below 1.0 per 1,000 km averaged over the most recent 50,000 km, demonstrated performance in all weather conditions present in the service zone (including the heavy rain, snow, and fog common in Seoul’s climate), and cybersecurity certification under the Korean Information Security Management System standard.

The Four Test Zones

Seoul’s autonomous vehicle testing ecosystem comprises four geographically distinct zones, each designed to present different operational challenges.

Sangam Digital Media City (DMC) Zone. The original and most mature test zone, established in 2020, covers 12.3 km of roads in the Sangam-dong commercial district of Mapo-gu. The zone includes four-lane arterials, commercial driveways, two signalised roundabouts, and the MBC/SBS/CJ ENM broadcast district’s pedestrian-heavy environment. The Sangam zone hosts the highest concentration of test vehicles — 28 AVs from seven operators — and has accumulated 1.8 million km of test data. The zone’s digital infrastructure includes 48 LiDAR-equipped roadside units (RSUs) that share real-time traffic data with connected vehicles through 5G C-V2X (Cellular Vehicle-to-Everything) communication, 120 smart traffic signals that broadcast signal-phase-and-timing (SPaT) data to approaching AVs, and 34 edge-computing nodes that process roadside sensor data and relay hazard warnings to connected vehicles with latency under 10 milliseconds.

Gangnam Autonomous Shuttle Zone. Established in 2023, this zone operates an autonomous shuttle service along a 4.2-km fixed route connecting Samseong Metro Station with the COEX convention complex, Bongeunsa Temple, and the Gangnam Finance Center. The shuttle — a six-passenger autonomous minibus manufactured by 42dot — operates at Level 4 without an onboard safety operator, monitored remotely from a control centre at 42dot’s Gangnam office. The service runs 14 hours per day (07:00-21:00), completes approximately 80 round trips daily, and has carried over 420,000 passengers since its launch. The route’s controlled environment — wide roads, moderate traffic, minimal pedestrian conflict — makes it an ideal proof-of-concept for autonomous public transit, though it does not replicate the complexity of general mixed-traffic operation.

Yeouido Autonomous Taxi Zone. Launched in late 2024, this zone permits autonomous ride-hailing service within a 6.8-km operating area covering Yeouido Island and adjacent sections of Yongsan-gu. Two operators — Kakao Mobility (using autonomous Hyundai IONIQ 5 vehicles) and 42dot (using its proprietary platform) — run a combined fleet of 12 autonomous taxis with safety drivers present but hands-off during normal operation. Passengers hail rides through the standard Kakao T and 42dot apps, and fares are set at 80 percent of conventional taxi rates. The zone serves approximately 800 rides per day and provides the first real-world dataset on consumer acceptance, pricing sensitivity, and operational economics of autonomous ride-hailing in a Korean urban context.

Cheonggyecheon Urban Navigation Zone. The most challenging of the four zones, established in 2025, covers a 14.7-km mixed-traffic corridor from Dongdaemun to Seoul Station via the Cheonggyecheon Stream road. This zone includes narrow lanes, heavy pedestrian crossing activity, bus rapid transit median lanes, delivery vehicle double-parking, and the concentrated complexity of Seoul’s central commercial district. No autonomous shuttle or taxi service operates here; the zone is exclusively for testing and data collection. The purpose is to build the operational datasets needed for eventual AV deployment in Seoul’s most demanding urban environments.

The Industrial Ecosystem: Korea’s AV Supply Chain

Seoul’s autonomous vehicle programme sits within a national industrial strategy that views AV technology as a successor export industry to Korea’s established automotive sector. The key corporate participants:

Hyundai Motor Group / 42dot. Hyundai acquired the AV startup 42dot in November 2022 for approximately KRW 400 billion, integrating 42dot’s software platform with Hyundai’s vehicle manufacturing capabilities. The combined entity’s AV programme — branded “PBV” (Purpose-Built Vehicle) — targets commercial deployment of Level 4 autonomous shuttle buses, delivery vehicles, and ride-hailing vehicles by 2028. Hyundai’s Ulsan factory is tooling a dedicated AV production line with an initial capacity of 3,000 units per year.

Samsung Electronics. Samsung’s AV involvement focuses on the sensor and computing hardware layer: LiDAR chipsets, neural-processing units (NPUs) for onboard perception computing, and 5G communication modules for V2X connectivity. Samsung’s Harman subsidiary provides the infotainment and vehicle-computing platform used by several AV test operators.

Naver Labs. Naver’s autonomous driving subsidiary operates three test vehicles in the Sangam zone and is developing a high-definition mapping platform — ARC (A Real-time City) — that provides centimetre-accuracy maps of Seoul’s road network updated in real-time from fleet sensor data. The ARC platform is positioned as infrastructure-as-a-service for third-party AV operators, generating revenue from map-data licensing rather than vehicle operation.

KT Corporation. Korea’s largest telecommunications company provides the 5G-based C-V2X communication infrastructure that connects roadside units with autonomous vehicles across all four test zones. KT’s network engineering for the Seoul AV programme has generated intellectual property in ultra-reliable low-latency communication (URLLC) that the company is licensing to international smart-city projects.

Safety Performance and Incident Analysis

The 4.2 million km of accumulated test data have produced a safety record that, while imperfect, demonstrates progressive improvement. Key metrics as of early 2026:

Total reportable incidents (defined as any event requiring safety-driver intervention to prevent a collision): 847 across all operators, yielding an incident rate of 0.20 per 1,000 km. Disengagement rate (safety-driver takeovers for any reason): 2.1 per 1,000 km across all operators, with the best-performing operator (42dot) achieving 0.8 per 1,000 km. Collision incidents: 14 total, of which 11 were classified as “minor contact” (less than KRW 500,000 in damage, no injuries) and three as “moderate” (property damage exceeding KRW 500,000, minor injuries to vehicle occupants). Zero fatalities or serious injuries have occurred in the Seoul AV test programme.

The incident pattern reveals consistent categories. The most frequent disengagement cause (38 percent of events) is unexpected pedestrian behaviour — jaywalking, sudden sidewalk-to-road entry, or pedestrians walking in the roadway while distracted by mobile phones. The second cause (22 percent) is delivery-vehicle obstruction — double-parked delivery trucks that block AV sensor lines and create narrow gaps requiring human judgment to navigate. The third cause (16 percent) is construction-zone navigation, where temporary road markings, barriers, and flaggers create operating conditions outside the AV’s trained domain.

These incident patterns reflect Seoul-specific urban characteristics that differ from the controlled environments where most Western AV testing occurs. Seoul’s pedestrian density (18,400 pedestrians per km of road per day in central districts), its delivery-vehicle culture (an estimated 890,000 daily parcel deliveries to Seoul addresses), and its frequent construction activity create a complexity profile that makes Seoul one of the most demanding AV operating environments in the world — and therefore one of the most valuable for technology development.

The Regulatory Sandbox: Balancing Innovation and Safety

Korea’s approach to AV regulation follows what MOLIT terms the “regulatory sandbox” model: permissive testing rules within defined zones, combined with rigorous safety-certification requirements for any expansion beyond sandbox boundaries. The sandbox model aims to avoid both the paralysis of over-regulation (which would push AV development to more permissive jurisdictions) and the risk of under-regulation (which could produce safety incidents that destroy public trust).

The sandbox has proven effective for testing but creates a scale challenge for commercialisation. The current Pilot Operation Permit structure limits commercial AV services to zones not exceeding 15 km of road network per permit. This is adequate for shuttle routes and small ride-hailing zones but insufficient for city-wide autonomous taxi operations or autonomous freight delivery at commercial scale. MOLIT’s proposed 2027 amendment to the Autonomous Vehicle Act would introduce a new “Urban Operation Domain” permit category covering entire districts (gu) rather than defined road networks, enabling AV operations across all roads within a district’s boundary once the operator demonstrates adequate safety performance.

The timeline for regulatory evolution reflects a deliberate progression: 2020-2025 (sandbox testing), 2025-2028 (limited commercial operations in defined zones), 2028-2032 (district-level commercial operations with safety-operator flexibility), and 2032-forward (full autonomous operation without safety operators in certified ODDs). This ten-to-twelve-year regulatory pathway is broadly consistent with the timelines projected by Hyundai Motor, 42dot, and the other Korean AV developers.

Autonomous Public Transit: The Bus and Shuttle Applications

The most immediate commercial application for autonomous vehicles in Seoul is not the robotaxi but the autonomous shuttle bus. Two factors drive this assessment. First, the economic case: Seoul’s bus operator workforce faces a demographic crisis, with an average driver age of 57.3 years and projected shortfalls of 2,400 drivers by 2028. Autonomous buses address a labour shortage, not merely a cost-optimization opportunity. Second, the operational case: fixed-route bus operations in median bus lanes present a more tractable AV problem than unconstrained ride-hailing, because the operating environment is controlled (separated from mixed traffic) and the route is predetermined.

The Sangam-to-Cheonggyecheon corridor pilot, operational since October 2025, deploys two Hyundai autonomous buses on a 6.3-km BRT route with safety drivers present. The buses operate at Level 3+ (approaching Level 4) in the median bus lane, with the safety driver monitoring but not controlling the vehicle during normal operation. The pilot’s primary objective is collecting the operational data — handling of bus-stop passenger boarding and alighting, interaction with median-lane turning vehicles, and response to pedestrians crossing the median lane — needed to develop the safety case for driverless BRT operations.

The Seoul Metropolitan Government’s transport planning division projects that autonomous operation of 15-20 percent of the metropolitan bus fleet (approximately 1,100-1,500 vehicles) is achievable by 2035, focusing initially on BRT corridors and late-night routes where driver availability is most constrained. The estimated operating cost savings — KRW 380-520 billion annually, primarily from driver labour elimination on converted routes — would substantially reduce the bus subsidy burden on the metropolitan budget.

Autonomous Freight and Last-Mile Delivery

The autonomous delivery vehicle market in Seoul is developing in parallel with passenger applications. Three operators — Naver Labs, Woowa Brothers (the Baemin food-delivery platform), and CJ Logistics — hold permits for autonomous delivery-robot testing on Seoul sidewalks and pedestrian zones. These small (sub-100-kg) robots operate at walking speed (4-6 km/h) on sidewalks, delivering food, parcels, and groceries within 2-km radii of commercial hubs.

Naver’s delivery robots have completed over 200,000 deliveries from the company’s headquarters campus in Bundang (Gyeonggi Province) and are being deployed in the Yeouido office district on a pilot basis. Woowa Brothers’ robots operate from 34 restaurant clusters across six Seoul districts, completing approximately 1,400 deliveries per day with a 96 percent on-time completion rate.

The larger autonomous freight opportunity involves full-size delivery trucks operating on expressways between logistics centres and urban distribution hubs. CJ Logistics — Korea’s largest logistics company — has been testing a Level 4 autonomous truck on the Gyeongbu Expressway (Seoul-Busan) corridor since 2024, with safety-driver supervision. The commercial case for autonomous trucking is driven by the same demographic pressures affecting bus drivers: Korea’s trucking workforce has an average age of 54 and faces projected shortfalls that threaten supply-chain reliability within the decade.

Infrastructure Requirements for Scaled AV Deployment

Scaled autonomous vehicle deployment in Seoul requires infrastructure investments beyond the vehicles themselves. The Seoul Metropolitan Government’s AV Infrastructure Readiness Plan identifies four priority areas:

5G C-V2X Network Coverage. Current coverage spans the four test zones (128 km). Expansion to all 8,190 km of Seoul’s classified road network would require approximately 12,000 additional roadside units at a total investment of KRW 480 billion. KT Corporation’s deployment roadmap targets 60 percent arterial-road coverage by 2030 and 90 percent by 2033.

High-Definition Mapping. Centimetre-accuracy 3D maps of the entire Seoul road network — including lane markings, kerb profiles, traffic signal positions, and permanent infrastructure features — require initial survey investment of approximately KRW 120 billion and ongoing updating costs of KRW 18 billion per year as the urban environment changes through construction, renovation, and seasonal variation.

Traffic Signal Upgrade. Of Seoul’s 12,400 signalised intersections, only 2,300 currently broadcast SPaT data compatible with connected AV systems. Upgrading the remaining 10,100 intersections requires hardware retrofit costs of approximately KRW 340 billion spread over a 2026-2032 implementation period.

Cybersecurity Infrastructure. The connected nature of autonomous vehicles creates a cybersecurity attack surface that extends to the entire urban transport network. The National Intelligence Service and MOLIT have jointly developed the Autonomous Vehicle Cybersecurity Framework, requiring all AV operators to implement intrusion-detection systems, encrypted V2X communications, and real-time anomaly monitoring. The infrastructure-level cybersecurity investment for Seoul is estimated at KRW 86 billion.

Public Perception and Social Acceptance

The Seoul Institute’s 2025 survey on autonomous vehicle acceptance found that 43 percent of Seoul residents would be “willing to use” an autonomous taxi or shuttle, 31 percent were “uncertain,” and 26 percent were “unwilling.” Among the willing cohort, the primary motivations were convenience (68 percent), curiosity (54 percent), and expected cost savings (41 percent). Among the unwilling, concerns about safety (82 percent), loss of human control (61 percent), and technology failure (57 percent) dominated.

The age distribution of acceptance mirrors international patterns: 58 percent of respondents aged 20-34 expressed willingness, compared to 38 percent of those aged 35-54 and only 21 percent of those over 55. The aging population’s lower acceptance rate is problematic because elderly residents — who face mobility limitations from age-related driving impairment — are precisely the demographic that autonomous mobility could most benefit. The Seoul Metropolitan Government’s AV public engagement programme includes targeted outreach to elderly community centres, offering supervised AV rides and information sessions designed to build familiarity and trust.

Outlook Through 2030 and Beyond

The 2030 horizon will see Seoul transition from AV testing to limited commercial operations. The most probable deployment sequence: autonomous shuttles on fixed BRT routes (2027-2028), autonomous taxis in defined zones expanding from Yeouido to Gangnam and Jongno (2028-2030), autonomous delivery robots at commercial scale across central Seoul (2028-2029), and autonomous freight trucks on expressway corridors (2029-2031).

Full city-wide autonomous vehicle operations — the scenario where a meaningful fraction of Seoul’s vehicle fleet operates without human drivers — remains a 2032-2035 proposition at the earliest, contingent on regulatory evolution, infrastructure buildout, and demonstrated safety performance that satisfies both MOLIT’s technical standards and the Korean public’s risk tolerance.

The prize for getting this right is significant. KOTI estimates that widespread autonomous vehicle deployment could reduce Seoul’s traffic fatalities by 60-80 percent (from the current 281 annual deaths to 56-112), recover KRW 3.2 trillion annually in congestion costs through optimised traffic flow, and reduce transport-sector CO2 emissions by 18-24 percent through smoother driving profiles and reduced deadheading. These are projections, not certainties. But they define the scale of the opportunity that Seoul’s autonomous vehicle programme is building toward.