Physical AI is moving from research into early commercial deployment in robotics and other embodied systems, but regulatory frameworks and liability questions remain unresolved, according to new analysis from GlobalData.
In a media release summarising its latest Strategic Intelligence report, GlobalData said physical AI represents a shift from machines that follow pre-programmed instructions to systems able to perceive, reason and act autonomously. The firm said the technology can be deployed in robots, autonomous vehicles, drones and industrial systems, with commercialisation beginning in 2026 across sectors including manufacturing, healthcare, mining, energy and transport.
GlobalData described physical AI as a combination of technologies including the Internet of Things, generative and agentic AI, machine learning, “world models”, vision-language models and vision-language action models. The company said physical AI systems typically operate in a continuous loop of perception, reasoning and context, learning and action, relying on local feedback and iterative adaptation rather than “recursive learning”.
William Rojas, Research Director in GlobalData’s Strategic Intelligence unit, said the shift involves AI moving beyond purely digital applications into systems that interact with the physical world and learn through experience. He said physical AI “gains skills via real-world experience, modulating force and movement”, which in turn affects control models.
GlobalData’s commentary also highlighted geopolitical and industrial competition around the technology. It said Japan, China, South Korea, Taiwan, Singapore and the US are leading in R&D investment and deployment, and pointed to Japan’s commitment of JPY10 trillion (US$63 billion) to advanced robotics and AI.
The release also referenced Nvidia forming a coalition with 10 companies and organisations with robotics expertise, and said China is coordinating funding and support across central and local governments. GlobalData said China’s 15th Five-Year Plan includes commitments to develop the technology and establish funding and risk-sharing mechanisms.
Rojas said the competitive push is being led out of Asia-Pacific, arguing that established precision robotics capability and state-backed investment have given the region an early advantage.
Alongside competition, GlobalData flagged governance challenges. It said AI regulation is “not well developed” and is struggling to keep pace with technical advances, with physical AI regulated via overlapping existing regimes rather than dedicated laws. The firm said technical standards are being used to bridge gaps, while liability remains an open issue.
Nilesh Raghoo, Associate Analyst in GlobalData’s Strategic Intelligence team, said regulation varies by region and that no specific physical AI law currently exists. Instead, he said machinery, product safety, product liability and horizontal AI laws can all apply at once, with much of the work focused on harmonising these requirements.
GlobalData’s report, titled “Physical AI”, focuses on geographic trends in the technology with an emphasis on China, Japan and South Korea, and includes industry case studies.
