Anthropic’s Claude artificial intelligence model has successfully planned a navigation route on Mars for NASA’s Perseverance rover, marking a rare real-world deployment of generative AI in interplanetary exploration.
The Jet Propulsion Laboratory of NASA enabled Claude to plan a path for the rover, which Perseverance followed for about 400 metres on the Martian surface last month. The path was checked and verified by JPL engineers before it was sent to the rover.
According to Anthropic, Claude used vision-based analysis of images from NASA’s HiRISE camera and digital elevation models to identify terrain features and generate a continuous path with waypoints. The AI then produced rover commands using Rover Markup Language, a specialised format used by NASA systems.
JPL engineers simulated the AI-generated route, analysing more than 500,000 telemetry variables to ensure the rover’s safety. Only minor adjustments were required, mainly due to details visible in ground-level images that Claude had not accessed.
The Perseverance mission succeeded in executing all AI-designated Martian routes on both the 1,707th and 1,709th Martian days, which correspond to the Earth dates of December 8 and December 10 in 2025.
NASA Administrator Jared Isaacman explained that the test demonstrated how autonomous technology could improve mission efficiency while increasing scientific results for deep space missions.
According to Anthropic, JPL engineers believe that Claude could reduce the time it takes to design routes by about half, although no specific numbers were given. The success of the test indicates that such AI systems may have an increasing role in future robotic missions.