A 43-year-old insurance company employee in South Korea started receiving threatening text messages last month. The sender, a customer disputing a traffic accident payout, warned the claims handler would be a "forensic target" for the Financial Supervisory Service's Special Investigation Bureau - language that turned out to be a patchwork of outdated regulations and AI-generated errors. It was one of over 200 near-identical messages sent by the same customer, according to the employee's firm. The incident reflects a sharp rise in malicious AI-generated complaints flooding insurers, police, and local government offices as tools like ChatGPT make it simple to produce professional-sounding grievances laced with fake information.
Industry data confirms the scale. General Insurance Association of Korea and Korea Life Insurance Association figures show 15,996 complaints were filed with insurers in the first quarter of this year, a 19.3% increase from 13,410 in the same period last year. The portion of complaints that used AI jumped from 5% in 2024 to 30% in the first quarter. An industry insider said, "When we run threatening messages or malicious complaints sent by customers through an AI detector, about nine times out of ten they are confirmed to have been written by AI."
Fake precision, real workload
Kim, the insurance employee on the receiving end, said, "At first, I thought a large law firm or professional damage adjuster was involved and became nervous. But upon closer examination, it was riddled with typical AI errors, such as patching together outdated regulations that had already been revised." The customer copied and pasted the same AI-generated text across hundreds of messages. Kim added that such customers are no longer rare.
The fabricated complaints often cite nonexistent departments or misuse technical jargon, forcing claims teams to spend time verifying each claim's factual basis. What appears authoritative is frequently hollow, built from legal phrases and agency names that sound plausible but don't align with current regulations.
Beyond insurance: police and civil servants overwhelmed
Police stations and city halls are dealing with the same flood. A 58-year-old public official in Hwaseong City, Gyeonggi Province, said, "In my 30 years as a public servant, I've always struggled with complaints, but it's never been this bad. The complaints are filled with similar phrases that even the complainants themselves don't understand." He warned that genuine cases needing government help are being buried under what he called "AI malicious complaints."
A police officer at a Seoul station said elderly residents are filing more criminal complaints because AI has made the process easier. "Investigators are already overwhelmed with cases, and these error-ridden 'AI complaints' are making things even harder," the officer said.
Why this matters for insurance professionals
Claims handlers, underwriters, and compliance teams now face a new layer of administrative friction: distinguishing legitimate customer grievances from AI-generated noise. Every flagged message that arrives loaded with pseudo-legal threats consumes time that could go toward actual claims resolution. The 30% AI-share figure suggests that nearly one in three complaints now requires additional scrutiny for factual accuracy, stretching teams that are already measured on speed and accuracy.
Understanding how generative AI produces these complaints - and what errors to look for - is becoming a necessary skill. Professionals who want to stay ahead of the challenge can explore resources like AI for Insurance Courses and ChatGPT Courses to learn how these tools work and how their output can be spotted. Recognizing the signatures of AI-written text, such as outdated legal references or misapplied jargon, is a practical defense against wasted investigative effort.
Your membership also unlocks:
