NASA has taken a significant step in applying artificial intelligence to space exploration by using an AI model to help plan driving routes for its Perseverance rover on Mars. In December 2025, engineers at NASA’s Jet Propulsion Laboratory relied on Claude, an AI system developed by Anthropic, to generate navigation commands for the rover. The effort marked the first time an AI system authored commands that were transmitted to and executed on another planet.
The test took place on December 8 and 10, corresponding to Martian days, or sols, 1707 and 1709. Claude was tasked with planning an approximately 400 meter route through a rocky area of the Martian surface. While modest in distance, the drive represented a milestone in reducing human workload for rover operations.
Because it takes about 20 minutes for signals to travel between Earth and Mars, rover operators cannot steer Perseverance in real time. Instead, they must plan routes in advance, defining precise waypoints that the rover follows autonomously. Until now, that process has been handled entirely by human experts.
How Claude Planned the Rover’s Route
Perseverance, which landed on Mars in February 2021, is designed to study the planet’s geology and past climate and to collect samples for eventual return to Earth. Its landing site, Jezero Crater, was selected due to evidence that it once held water and may have supported microbial life.
Driving across the Martian surface is risky. In 2009, NASA’s Spirit rover became trapped in soft sand and was unable to continue its mission. To reduce similar risks, JPL engineers carefully plan each drive using orbital imagery and data from the rover’s onboard cameras.
For this test, engineers provided Claude with operational data, images, and years of accumulated driving experience. The AI used this information to generate waypoints using Rover Markup Language, an XML-based programming language developed for earlier Mars missions. Claude analyzed overhead images, broke the route into 10 meter segments, and iteratively refined the path by reviewing and revising its own output.
Before sending the commands to Mars, engineers ran the AI-generated plan through Perseverance’s standard simulation system. More than 500,000 variables were modeled to predict the rover’s position and identify potential hazards. Human reviewers made only minor adjustments based on ground-level imagery before approving the route. Perseverance then successfully completed the drive.
Implications for Future Missions
NASA engineers estimate that using AI-assisted planning could cut route preparation time by about half. Faster planning could allow for more frequent drives, increased data collection, and reduced training demands for new operators.
The test also offers insight into how AI could support future missions with greater autonomy. NASA’s Artemis program, which aims to return astronauts to the Moon and establish a sustained presence near the lunar south pole, is expected to rely heavily on automated systems. Longer-term missions to Mars and beyond may depend even more on AI as communication delays increase and operating environments become more challenging.
For now, the Perseverance experiment shows how general-purpose AI tools are beginning to move from Earth-based applications into critical roles in space exploration.
