Kakao Mobility isn’t just dreaming about driverless cars. The South Korean mobility giant has laid out a detailed, in-house plan to achieve Level 4 autonomous driving as part of its broader physical AI strategy. Vice President Kim Jin-kyu, who heads the company’s Physical AI division, delivered the roadmap at the 2026 World IT Show in Seoul. The event, held under the banner Beyond Idea, Into Action: AI moves Reality, drew 460 organizations from 17 countries. South Korea’s Ministry of Science and ICT framed the conference as part of a larger shift toward physical AI, where artificial intelligence moves beyond screens and into factories, roads, and infrastructure.
Kim explained that Kakao Mobility is working to stitch autonomous driving technology directly into Korea’s physical infrastructure. The company isn’t just building a better self-driving car. It wants to create an open ecosystem that helps local players stay competitive. At the heart of this push is Level 4 autonomy, a designation from the U.S. National Highway Traffic Safety Administration for systems that can handle all driving tasks in limited service areas. Think autonomous taxi zones or fixed downtown districts. The passenger becomes a passenger, not a backup driver.
Three Pillars of the Level 4 Roadmap
Kakao Mobility’s Level 4 strategy rests on three technical pillars: machine learning models, vehicle redundancy, and validation systems. The company is developing ML models that handle perception, decision-making, and control without human input. That means the vehicle reads its surroundings, decides what to do, and executes maneuvers entirely on its own. No joystick. No remote operator tapping in for tricky turns.
The second pillar is redundancy. Kakao Mobility plans to use vehicle architectures where critical systems have backups. If a sensor fails or a brake line goes soft, a secondary system takes over. It’s the automotive equivalent of a pilot having a co-pilot. The third pillar is a validation platform that blends virtual simulations with real-world driving data. This isn’t just a testing tool. It’s designed to accelerate performance improvements and quality checks as the company rolls out commercial services.
Safety and the Human Factor
Autonomous vehicles are only as good as their safety nets. Kakao Mobility is building an integrated safety management platform, and one piece of it is particularly clever: the Autonomous Vehicle Visualizer. This is a 3D visualization tool that shares the vehicle’s field of view in real time. Passengers can see exactly what the car is detecting, from pedestrians to potholes. It turns a potentially unnerving ride into something transparent, even reassuring. You might call it a trust dashboard.
The company also plans to add a 24-hour control center and an anomaly detection system powered by vision-language models. These models analyze visual data and contextual cues simultaneously. If the car encounters something strange, a wrong-way driver or an unexpected construction zone, the system can flag it for remote intervention. Kakao Mobility hasn’t disclosed the model architecture or performance benchmarks for this system yet. But the intent is clear: autonomous driving doesn’t mean unattended driving. Someone, somewhere, is paying attention.
Opening the Toolbox: An Open Ecosystem Strategy
Kakao Mobility wants to be a platform, not just a service provider. The company plans to share selected technology assets with other companies, startups, and manufacturers. These assets include large-scale autonomous driving datasets, high-definition (HD) maps, and platform APIs for ride-hailing and dispatch. HD maps are especially critical. They provide the centimeter-level detail that autonomous vehicles need to localize themselves and make driving decisions. Without HD maps, a self-driving car is flying blind in the literal sense.
The idea is to let other industry participants develop their own autonomous technologies without rebuilding the entire stack from scratch. Kakao Mobility will also share operational resources, like fleet management systems and on-site response capabilities. It’s a bet that a rising tide lifts all cars. By lowering the barrier to entry, the company hopes to foster a domestic ecosystem that can take on global players like Waymo and Baidu. Whether that gamble pays off depends on execution, but the strategy is sound: compete by enabling competition.
Real World Results: The Gangnam Late-Night Service
Kakao Mobility isn’t just talking. The company pointed to its late-night autonomous vehicle service in Seoul’s Gangnam district as a proof of concept. Available through the Kakao T platform, the service lets users hail a self-driving taxi alongside regular cabs. From its launch on September 26, 2024, to February 28, 2026, the fleet logged 7,754 rides, according to the Seoul Metropolitan Government. The city reported zero accidents attributed to autonomous driving technology during that window. The service averaged about 24 trips per operating day.
In April 2026, the service moved from a free pilot to paid operation. Seoul also expanded the fleet from three vehicles to seven, plus two reserves. Users can access the service through the Kakao T app via a dedicated Seoul Autonomous Car icon or through the regular taxi-hailing menu. Kakao T bundles multiple mobility services in one app, from navigation to vehicle maintenance. The Gangnam service shows that Level 4 isn’t just a lab experiment. It’s already carrying passengers through one of Seoul’s busiest districts, late at night, when traffic is lighter but stakes are high.
This is where the rubber meets the road, literally. The data from Gangnam will feed back into Kakao Mobility’s machine learning models and validation systems, creating a virtuous cycle of improvement. The more miles the fleet logs, the smarter the platform becomes. And the smarter the platform becomes, the safer and more efficient the service gets.
Looking ahead, the real question isn’t whether Kakao Mobility can build Level 4 technology. It’s whether the company can turn a promising pilot into a scalable business. The open ecosystem strategy suggests they recognize that no single player can own the entire autonomous driving stack. By sharing data, maps, and APIs, Kakao Mobility is betting that collaboration will outpace isolation. If they’re right, the roads of Seoul might look very different in five years. And the rest of the world will be watching.
