The Perseverance rover, active on Mars since 2021, has pioneered the use of artificial intelligence for navigational pathfinding. In December 2025, it successfully navigated approximately 400 meters across a section of the Jezero Crater, with the route planned by Anthropic’s chatbot, Claude.
Devising a path for a rover is inherently difficult, even for human experts. Numerous variables must be accounted for to prevent slips, tipping over, or the vehicle becoming immobilized. Typically, NASA ground controllers manually plot these trajectories, relying on imagery from orbital assets and the rover’s onboard cameras.
To train Claude Code, Anthropic’s agent, it required extensive contextual data amassed by Perseverance over several years. The model then systematically generated the traverse line by linking ten-meter segments and subjecting each segment to rigorous evaluation.
Engineers at NASA’s Jet Propulsion Laboratory (JPL) meticulously vetted the path suggested by Claude using specialized simulation software. Ultimately, only minor adjustments were needed, stemming from access to imagery that Claude lacked during its initial planning phase.
According to engineering estimates, deploying Claude is expected to halve the time currently dedicated to route planning and enhance the path’s overall consistency. This efficiency gain will enable rover operators to execute more traverses, amass greater volumes of scientific samples, and conduct broader analyses.
NASA harbors optimism that autonomous AI systems will facilitate deeper exploration into more distant regions of the Solar System by its probes. This capability is particularly crucial given NASA’s recent 20% staffing reduction and stringent budgetary constraints. Early in 2026, Congress vetoed a proposal that aimed to nearly halve NASA’s science budget, electing instead to maintain funding at the 2025 level.
About the Author
