Park Han-sol reports on Korea's financial regulators, along with fintech and insurance. She previously wrote about the art world, from biennales and exhibitions to fairs and auctions, with a focus on Seoul and the figures shaping the scene. Before joining The Korea Times, she spent a year at ABC News' Seoul bureau, contributing to coverage of major Asia-Pacific events.
AI-generated complaints create new headaches for insurers
Image generated by ChatGPT
By Park Han-sol
A recent customer complaint filed with one of the top three life insurers in Korea appeared convincing at first glance. It cited what seemed to be relevant laws and argued that the company had misinterpreted them.
There was one problem: The law did not exist.
"Sometimes customers cite statutes that simply aren't real," a company official said. "Other times the law itself exists, but the interpretation is completely off. We're seeing more grievances that sound plausible on the surface but don't hold up under scrutiny."
At another big-name non-life insurance firm, claims handlers encountered a different telltale sign of artificial intelligence (AI). One submission still contained the asterisks that ChatGPT typically uses for emphasis.
"People can put a few details into ChatGPT and ask it to draft a complaint for them. We are definitely seeing cases like that," an official at the insurance firm said.
These episodes illustrate a growing challenge for the country's insurance sector as more consumers turn to generative AI tools to draft complaints and appeals.
While chatbots can help explain complex policy terms and legal concepts, they remain prone to so-called hallucinations, generating fictitious statutes, fabricated precedents and inaccurate interpretations with a convincing degree of confidence.
Industry officials say many consumers submit such material without independently verifying it.
Yet insurers cannot simply dismiss these complaints because they appear to have been generated by AI. Once a case is filed, claims teams are required to review each allegation and provide a formal response.
"Regardless of how it's written, we still have to examine every claim that's raised," an official at a local life insurer said. "If the dispute involves a policy sold in 2020, we have to pull up the original files and verify everything line by line. That's where a lot of time and resources go."
Insurance companies say the growing use of AI is adding to administrative workloads. The spread of mobile services has already lowered barriers to filing claims and complaints, while generative AI tools can instantly produce lengthy submissions complete with legal arguments, regulatory references and purported precedents.
Korea's 39 insurers received a total of 15,996 complaints in the first quarter, up 19.3 percent from a year earlier. Payout disputes accounted for 65.4 percent of the total.
While firms cannot reliably determine how many filings were generated with AI, insiders note that the technology is contributing to longer and more complex submissions.
"These writings are getting much longer because customers are no longer drafting them from scratch," the non-life insurance official said. In one case, a policyholder submitted dozens of complaints over a matter of months involving essentially the same dispute, with each version only slightly reworded using AI.
Some observers argue that heightened regulatory scrutiny may be creating unintended incentives. As financial authorities place increasing emphasis on consumer protection and insurers become more sensitive to grievance statistics, some policyholders appear to believe that persistent complaints improve their chances of securing a payout.
Regulators are also watching for signs that generative AI could facilitate insurance scams.
Earlier this month, the Financial Services Commission announced it would develop countermeasures after identifying new schemes involving AI-generated medical records, including diagnostic reports and itemized treatment statements, to support fraudulent benefit claims.
The challenge is compounded by legal restrictions on access to medical information. Under Korea's Medical Service Act, medical records generally cannot be provided to third parties without a patient's consent, limiting insurance companies' ability to independently verify submitted documents.
The rise of AI-assisted schemes comes as the industry is already grappling with increasing levels of insurance fraud. According to the Financial Supervisory Service, the amount of detected private insurance fraud reached a record 1.16 trillion won ($767.3 million) last year.