NASA tests AI satellite that acts on its own in space; details inside

In a world increasingly driven by automation and AI, NASA is quietly working on something that could fundamentally change the way we observe our planet from space. The American space agency has begun testing a system that gives satellites the ability to make independent decisions– choosing what to observe, when to act, and how to respond, all without any human involvement.


It’s a shift that reflects the growing need for smarter, more efficient monitoring of Earth– especially at a time when climate-related events like wildfires, heatwaves, and extreme storms are becoming more frequent and harder to track in real-time. The new system, called Dynamic Targeting, is NASA’s step towards building a new generation of satellites that can think on the move.


This isn’t about adding more cameras in space. It’s about giving satellites a brain.

Traditionally, Earth-observing satellites have followed a rigid routine– pass over a fixed location, take images, and beam that data back to Earth for scientists to study. But that approach has its limits. One major issue is cloud cover, which blocks the satellite’s view. More than half the time, these expensive instruments end up collecting images that are largely unusable.


NASA’s new method looks to solve that. With the help of onboard AI, satellites can now scan the path ahead in real time, identify whether the view is clear or blocked, and decide whether or not to capture the image.


“If you can be smart about what you’re taking pictures of, then you only image the ground and skip the clouds. That way, you’re not storing, processing, and downloading all this imagery researchers really can’t use,” said Ben Smith of NASA’s Jet Propulsion Laboratory (JPL).

The first real test of this technology took place aboard a compact satellite called CogniSAT-6, launched in March 2024. Built by Open Cosmos and equipped with an AI chip from Ubotica, the satellite was given a new set of abilities– not just to observe, but to decide.


Since the satellite doesn’t have a separate sensor to scan ahead, it tilts forward by up to 50 degrees to get a preview of the upcoming terrain. An AI algorithm then checks whether the area is cloud-covered. If it’s clear, the satellite tilts back and captures the image.


This process– scanning, analysing, planning and capturing– takes less than 90 seconds, all while the satellite is moving at nearly 17,000 mph (7.5 kilometers per second).

What’s remarkable here isn’t just the technology, but how it’s being used to make satellites more selective and aware. The goal is to get them to interpret what they see, not just record it.


“The idea is to make the spacecraft act more like a human: Instead of just seeing data, it’s thinking about what the data shows and how to respond,” explained Steve Chien, Technical Fellow in AI at JPL and the principal investigator for Dynamic Targeting.

Right now, the AI is trained to detect and avoid clouds. But NASA has much bigger plans. The next phase of testing will involve satellites using similar decision-making to actively look for cloud formations linked to storms or climate events.


Eventually, the system will also be trained to spot heat signals and thermal anomalies– useful for identifying wildfires, volcanic activity, or rapidly changing weather patterns. Specialised algorithms will allow the satellite to assess each scene and respond accordingly.


“This initial deployment of Dynamic Targeting is a hugely important step,” said Chien. “The end goal is operational use on a science mission, making for a very agile instrument taking novel measurements.”