Insurance companies are using AI-powered computer models to analyze aerial images of homes and automatically flag properties for policy nonrenewal based on perceived risks, leading to customers being dropped without seeing the evidence against them.
Insurance companies across the U.S. are deploying aerial surveillance programs using drones, manned airplanes, and high-altitude balloons to photograph properties, with the industry-funded Geospatial Insurance Consortium covering 99% of the U.S. population. Computer models analyze these images to identify underwriting issues such as damaged roof shingles, yard debris, overhanging tree branches, and undeclared swimming pools or trampolines. The flagged images are used as evidence for policy nonrenewal notices. Cindy Picos from northern California was dropped by CSAA Insurance based on aerial photos showing her roof had 'lived its life expectancy,' though an independent inspection found the roof had another 10 years of life. Her insurer initially refused to let her see the images. State Farm announced it would drop coverage on around 30,000 residential and 42,000 commercial properties in California. Former Farmers Insurance agent Nichole Brink reported seeing customers dropped based on images that were two to three years old, including cases where tree shadows were misidentified as tree limbs. Consumer advocates report a dramatic increase in nonrenewal notices based on aerial images, with concerns about accuracy of image interpretation and limited customer rights to challenge decisions.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
AI systems that fail to perform reliably or effectively under varying conditions, exposing them to errors and failures that can have significant consequences, especially in critical applications or areas that require moral reasoning.
AI system
Due to a decision or action made by an AI system
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed