What happened
Maeil Business Newspaper reports that Autonomous A2G was selected for South Korea’s 2026 Investment and Loan-Linked Technology Development Project. The reported project is budgeted at about 6.6 billion won, including 5 billion won in government R&D support, and is scheduled to run through June 2030. A2G plans to develop a general-purpose physical-AI system for unmanned autonomous-driving services, combining vehicle technology with control-center and remote-operation capabilities. The company says it will develop the system for both left-hand-drive environments such as Korea and right-hand-drive environments such as Singapore and Japan.
Maeil Business Newspaper, also known as MK, reports that Autonomous A2G was selected on Aug. 24 for the “Global Tips General Type” track of South Korea’s 2026 Investment and Loan-Linked Technology Development Project. The program is organized and hosted by the Ministry of SMEs and Startups and the Small and Medium Business Technology Information Promotion Agency, according to the report. MK says A2G’s selected task is titled “Development of Physical AI Commercialization Technology for Global Export of Unmanned Self-Driving Services.” The report describes the selection as a government R&D award rather than a commercial launch or a completed autonomous-driving deployment.
MK reports that the project will cost about 6.6 billion won over four years, ending in June 2030, with 5 billion won in government R&D support. The report does not specify the timing of payments, the conditions attached to the support, or whether the remaining funding has already been secured. A2G’s stated objective is to develop a general-purpose autonomous-driving technology that can operate across different national road environments and traffic laws. The system would be combined with unmanned-operation technologies, including control and remote operation, to create what the company describes as a single unmanned Level 4 autonomous-driving service package.
The central technical plan, as described by MK, is to avoid building a separate autonomous-driving model for every country. Instead, A2G intends to use a common physical-AI model alongside safety and legal-response systems adapted to local traffic laws and driving cultures. The company plans to support both left-hand-drive environments, including Korea, and right-hand-drive environments, including Singapore and Japan. MK reports that the work will include a right-hand-drive data-collection and learning system, physical-AI models covering both steering configurations, and links among vehicles, control centers and remote-support systems.
The report also lists practical steps beyond model development. These include international authentication related to electromagnetic compatibility and cybersecurity, obtaining local permits for unmanned autonomous driving, and connecting exported services with local demand. MK reports that core technology development and commercialization will be based in Singapore, where A2G has pursued autonomous-driving demonstrations and commercialization under a right-hand-drive road environment. The company says it will use locally collected driving data and licensing experience there, then apply the service package to airports, ports, industrial complexes, tourist destinations, autonomous shuttles and public-transportation links. The report names Singapore, Japan and the UAE as existing target countries, alongside wider markets in Asia, the Pacific and the Middle East.
Read the primary source: mk.co.kr ↗
Why it matters
The reported award would give A2G a multi-year public-financed effort to address one of the practical barriers to autonomous driving: adapting systems to different traffic rules, road environments and driving conventions. The project also links AI model development with certification, cybersecurity, remote support and local operating permits. If the work produces a deployable package, it could support autonomous shuttles and services in airports, ports, industrial complexes and tourist areas. However, MK is the only source provided, and the report does not independently establish the project’s milestones, technical performance, funding schedule or commercial results.
The reported award matters because it places physical AI at the center of a public effort to move autonomous driving toward exportable services. In this context, physical AI refers to systems that perceive and act in the real world, where failures can affect passengers, pedestrians, vehicles and infrastructure. A four-year program could give A2G time to work on data, control, certification and operations together. That is materially different from announcing a model or publishing a benchmark, although the source does not provide evidence that the project has yet produced a new capability.
The cross-border problem is significant. Autonomous-driving systems are shaped by lane orientation, road layouts, signs, traffic behavior, vehicle configuration and local regulation. MK reports that A2G wants a common model to cover left- and right-hand-drive settings while adding country-specific safety and legal systems. If that separation works in practice, it could reduce duplicated development and make deployments easier to adapt. The report does not provide comparative test results, error rates or evidence showing that one model can perform equally well across the environments described.
The project also highlights that autonomous driving is not only an AI-model problem. The reported plan includes vehicle-to-control-center links, remote support, cybersecurity, electromagnetic certification and local permits. Those requirements determine how a system is supervised, how incidents are handled and whether an operator can legally run it without a person in the vehicle. They also create additional points of failure beyond perception and planning. MK does not report the proposed staffing model for remote operations, the safety case for Level 4 service, or the conditions under which human intervention would be required.
For public policy and industry, the award is a concrete example of government funding being used to support physical-AI commercialization and international expansion. It may provide a test of whether public R&D can help a smaller company move from demonstrations to repeatable services in regulated environments. At the same time, the evidence is limited. The only supplied source is MK’s translated report, and no government award notice, contract, independent technical assessment, customer agreement or permit record was provided for confirmation. The selection should therefore be treated as reported, while its eventual impact remains unestablished.
What to watch next
The key tests will be whether A2G can collect and use right-hand-drive data, demonstrate reliable performance across both traffic orientations, and secure the permits and certifications required in each target market. Watch for evidence from Singapore trials, independent safety assessments, cybersecurity reviews and clearly defined operating limits for remote support. It is also not yet clear how much of A2G’s technology is already deployed, how the reported government support will be disbursed, or whether expansion to Japan, the UAE and other markets has been formally contracted.
The first thing to watch is whether A2G publishes a detailed project schedule. Useful evidence would include milestones for right-hand-drive data collection, model training, vehicle-control integration and unmanned-driving readiness assessments. The source says the project runs through June 2030 but does not identify interim targets or acceptance criteria. Information about the government support’s payment schedule and the project’s private contribution would also clarify how much work is funded at each stage.
Singapore is likely to be the most informative early test because MK identifies it as the base for development and commercialization. Future reporting should distinguish between a demonstration, a safety evaluation and a permitted commercial service. Evidence should include the operating area, weather and traffic conditions, supervision arrangements, intervention rates and any restrictions placed on the vehicles. A company statement that a system is “Level 4” would not by itself establish that it has received approval for a particular public service.
The right- versus left-hand-drive claim also needs direct testing. Watch for results that compare the same system across Korea, Singapore or Japan, including failures at intersections, lane changes, unfamiliar road layouts and interactions with other road users. Independent testing would be more informative than internal demonstrations alone. Cybersecurity reviews, electromagnetic certification and records of local autonomous-driving permits will show whether the project’s operational and regulatory work is progressing alongside its AI development.
Finally, watch for evidence of actual demand. MK reports possible applications in airports, ports, industrial complexes, tourist destinations, autonomous shuttles and public-transport links, as well as expansion to Singapore, Japan, the UAE and other regions. It does not identify signed customers, deployment dates, vehicle counts, revenue or service contracts. Those details would help separate a funded development plan from a commercially established product. Until they emerge, the most meaningful unknowns are performance in varied environments, the reliability of remote operations, the scope of public funding and whether A2G can obtain permission to operate without onboard staff.
