Researchers have identified eight key risks in autonomous geoinformation systems

A group of scientists published a review study dedicated to the ethical and technical threats when implementing large language models in geoinformation systems (GeoAI). The authors analyzed the work of autonomous agents that use natural language to process spatial data and identified eight critical issues that require immediate regulation. The list of threats included passive location inference based on movement data, amplification of spatial distortions due to data autocorrelation and hallucinations when interpreting geographical facts. Researchers note that current protection methods used in general AI systems do not take into account the specifics of spatial data, where scaling errors and structural risks can lead to incorrect management decisions. To solve these problems, the authors proposed an autonomous GIS architecture based on the principles of risk-informed management. The system includes auditable artifacts and controlled mechanisms at each stage of the data life cycle. The effectiveness of the approach is demonstrated on a scenario of automated route planning during flood relief. The work highlights a significant gap between theoretical security concepts and their practical implementation. Most of the proposed protection methods are currently conceptual and require field testing. The study is published under the Creative Commons Attribution 4.0 license. The results of the study set a vector for future developments in the field of GeoAI, focusing on empirical validation and the creation of interpretability tools specific to spatial tasks. The implementation of such standards is necessary to prevent systemic errors in critical infrastructure.