Lawyers fear losing trust as courts consider mandatory AI disclosure
Korea’s judiciary is considering rules to require lawyers to disclose AI use in court filings, fueling concerns over trust, enforcement and the future of legal tech.
The Supreme Court building in Seocho District, southern SeoulNEWS1
Should lawyers have to disclose when AI helped write a court filing? As AI spreads into professions built on trust, the question is exposing the gap between calls for transparency and how technology is actually being used.
The National Court Administration (NCA) plans to require parties to disclose AI use in documents submitted to courts through a revision to Supreme Court rules, legal sources said Thursday.
While the administration discussed the proposal with the Korean Bar Association (KBA) last month, the association said the rules were unclear. The administration subsequently revised the proposal, asking whether disclosure should be required specifically when AI-generated answers are used to cite materials such as precedents, statutes, academic papers and legal commentaries.
The push follows a rise in court filings containing fabricated citations, according to the administration.
“Using AI to polish an argument and make it sound persuasive is a matter of choice, but falsely citing objective materials is a problem,” an official from the NCAsaid. “If AI was used, that use should be disclosed.”
In March, an administration task force proposed making litigants bear the costs of delays or unnecessary proceedings caused when they submit fabricated statutes or other false materials generated with AI.
The KBA opposes the proposal to disclose AI usage, arguing it would be difficult to implement without clearer standards.
“It is unclear whether lawyers would have to disclose AI use even when it was used only to prepare an initial draft,” the association said.
Behind the opposition lies a deeper worry: that an AI label could itself become a mark against a lawyer. One anonymous lawyer, who said AI is used in more than eight out of every 10 filings they prepare, questioned how judges might perceive such a disclosure.
“Wouldn’t a label saying AI was used make the filing appear less trustworthy to the court?” the lawyer said.
The Korean Bar Association building in Seocho District, southern SeoulNEWS1
Similar concerns have emerged in other areas where AI use is disclosed. TikTok posts labeled AI-generated or AI-edited received about 7 percent to 8 percent fewer likes than posts without the label with the same number of views, according to a study published in May by researchers at the University of Southern California’s Marshall School of Business. The study examined 1.14 million TikTok posts in the United States. Overall engagement, including likes, comments and shares, was about 7 percent lower. The researchers attributed the decline to perceptions that disclosing AI use signals reduced effort and lower credibility.
Legal technology companies also warn that blanket disclosure requirements could dampen Korea’s still-developing legal AI market.
“Disclosing AI use follows the same principle as citing sources in academic papers and reports,” said Lee Seong-yeob, a professor at Korea University’s Graduate School of Management of Technology. “But recent legal AI services have significantly reduced hallucinations, and trying to solve early-stage problems through regulation could mean losing the benefits AI provides.”
The debate extends well beyond the legal profession. Similar concerns are spreading across scientific and other expert communities.
At the International Conference on Machine Learning, held in Seoul from July 6 to 11, organizers rejected 497 papers submitted by authors who had used AI to conduct peer reviews of other researchers’ work.
“We hope that by taking strong action against violations of agreed-upon policy we will remind the community that as our field changes rapidly the thing we must protect most actively is our trust in each other,” the conference said in a blog post explaining the decision.
Choi Byoung-ho, a professor at Korea University’s AI Research Institute, said requiring disclosure for every use of AI would be overly broad and difficult to enforce.
“Requiring an AI label for every instance of AI use would amount to sweeping regulation, and there is no practical way to verify compliance,” Choi said. “A more realistic approach would be to impose fines in cases where AI hallucinations cause actual harm.”
This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.
