Dartmouth researcher Sean J. Westwood has developed an AI bot that completes surveys for five cents each while passing 99.8% of standard fraud detection measures. The bot can manipulate polling results with minimal effort – just 10 to 52 fake responses could flip which candidate leads in election polls. In one test, a single command reduced mentions of China as America’s top military rival from 86.3% to 11.7%. The AI system maintains consistent demographic personas and shows human-like reasoning patterns, making detection nearly impossible with current methods. Research Defender estimates 31% of raw survey responses already contain fraud. Westwood warns that unlike random noise from traditional fraud, AI-generated responses introduce targeted bias that could corrupt public health studies, political polling, and policy research that relies on understanding genuine public opinion. (Story URL)
