A research team at Wuhan University (China) has granted AI technology full control over a satellite, allowing it to operate independently for 24 hours.
Chinese researchers have reported that an artificial intelligence (AI) system was temporarily given complete control over a satellite in low Earth orbit as part of a groundbreaking experiment to test this technology’s capabilities in space.
AI has controlled a small Chinese satellite to observe locations in India and Japan. (Illustration: Shutterstock).
According to the journal Geomatics and Information Science from Wuhan University, over the course of 24 hours, the Earth observation satellite Qimingxing 1 was directed by an AI on the ground without any commands, assignments, or human intervention.
Scientists stated that the AI selected several locations on Earth and instructed Qimingxing 1 to examine them in greater detail. No explanation was provided regarding why the AI chose those specific sites.
One target area was the ancient city of Patna, located along the Ganges River in northeastern India. This city is also home to the Bihar Regiment, an Indian Army unit that clashed with Chinese soldiers in the Galwan Valley during a border dispute in 2020.
Along with Patna, Osaka, one of Japan’s busiest ports, where U.S. Navy ships occasionally operate in the Pacific, was also highly prioritized in the AI’s interests.
“This approach breaks existing rules for mission planning,” stated researcher Wang Mi and colleagues at Wuhan University.
So far, most satellites require specific commands or missions before taking action. These missions may arise in unexpected situations, such as during wars or earthquakes, or a satellite may be scheduled for long-term observational tasks regarding specific targets.
According to the research team, while AI technology is increasingly used in space missions—including image recognition, trajectory plotting, and collision avoidance—it has not been granted control over satellites, leading to wasted time and resources.
China currently has over 260 remote sensing satellites in orbit, but they often operate ineffectively, collecting low-value data that is time-sensitive without any specific missions.
Satellites are costly and have a limited lifespan. Therefore, researchers argue that maximizing their value with new orbital applications is urgent.
The team proposed that an AI-controlled satellite could alert users—including military forces, national security agencies, and other relevant bodies—if it detects unusual objects or activities.
However, to make correct decisions, AI needs a comprehensive understanding of Earth. Thus, it must not only identify artificial and natural objects but also learn about the complex and evolving relationships between them and different human societies.
Wang’s team has built a massive text data library from around the world, similar to the datasets used to train language models like ChatGPT. The AI developed by the Chinese researchers cannot engage in conversation, but it can operate proactively based on its training and increasing understanding of natural and human activities.
Researchers noted that the decision-making process of AI is exceedingly complex. The machine must weigh numerous factors—such as real-time cloud conditions, camera angles, target values, and the satellite’s mobility limits—when planning a workday.
They suggested that AI’s calculations could be sophisticated enough that traditional computer models for managing space missions might not solve the problem within a reasonable timeframe.
SpaceX’s Starlink satellite. (Photo: Space).
Elon Musk’s SpaceX has launched over 3,000 Starlink devices, while China plans to launch nearly 13,000 communication satellites to compete.
The U.S. and Chinese militaries have also proposed adding monitoring tools and additional features that allow small orbital spacecraft to do more than just provide internet services.
Thousands of satellites in the networks of Beijing and Washington will pose serious challenges for most space agencies. Some scientists argue that only AI could manage them without causing mass collision incidents.
Chinese space scientists pioneered the application of AI technology in orbit in 2021 by using a smart small satellite to detect and monitor a classified U.S. naval exercise off the coast of New York.
A space scientist in Shanghai, not involved in this research, expressed that they are not concerned about the prospect of granting AI greater control over satellites.
“I believe that Beijing will monitor closely. If AI attempts anything unusual, it will be immediately cut off by the operator,” the unnamed researcher stated.
However, another space scientist in Beijing, with experience in ChatGPT, expressed concerns about what might happen if a large language model gained access to flight control and learned how to operate satellites.
“Satellites are our eyes and ears in the sky. Do we really want AI to decide what we see and hear?” the expert remarked.
When asked what they would do if they were satellite operators on a day without a mission, ChatGPT mentioned conducting maintenance checks, software and firmware updates, calibrations, and tests, as well as exploring new areas of interest.
ChatGPT, developed by the OpenAI company backed by Microsoft, named several locations it is most interested in, including Australia’s Great Barrier Reef, the Amazon rainforest, the Himalayas, the Sahara Desert, and the Antarctic Peninsula.
When asked why it chose these areas, ChatGPT explained that they are affected by human activity and climate change.
