Blab AI
AI-Generated Audio Fuels Defamation Case Against South Korean Actor Kim Soo-hyun
In March 2025, YouTuber Kim Se‑eui released an audio clip that purportingly featured the voice of late actress Kim Sae‑ron, alleging that actor Kim Soo‑hyun had a sexual relationship with her. The recording was later confirmed to be fabricated using artificial intelligence (AI). The case illustrates how synthetic media can accelerate the spread of misinformation and complicate legal verification.
The audio, posted on March 10, 2025, claimed that Kim Soo‑hyun and Kim Sae‑ron had dated from middle school to college and that their first sexual encounter occurred during a winter break in Kim Sae‑ron’s second year of middle school. The clip was used by Kim Se‑eui to accuse Kim Soo‑hyun of defamation. Within eight days, 2,026 news outlets reported on the allegation, prompting Kim Soo‑hyun to hold a press conference in which he denied the claims.
Police investigation revealed that the file submitted for forensic analysis was not the original recording. The National Forensic Service stated it could not determine whether the audio had been manipulated because the file was incomplete. However, investigators concluded that the recording was fabricated based on several factors: multiple conflicting versions of the clip circulated, the distribution trail suggested manipulation, and KakaoTalk messages linked to the case appeared altered.
Attorney Ko Sang‑rok, representing Kim Soo‑hyun, said investigators considered the existence of “completely different versions of a recording said to capture the same conversation on the same day and at the same time” distributed to multiple parties. Former prosecutor Lee Chang‑hyun noted that while proving fabrication in court is challenging, advances in analytical technology may eventually allow reliable detection of AI‑generated evidence.
Professor Yoo Hyun‑jae of Sogang University warned that the case demonstrates how AI‑generated material can be weaponised for defamation. He said the incident “has lifted the lid on Pandora’s box” and that without institutional safeguards, more individuals could use similar techniques to influence public opinion.
The case has attracted attention from law enforcement and the media. Kim Se‑eui was arrested and sent to prosecutors on charges of defamation and using AI‑manipulated audio. His real estate was reportedly locked by court order, and the case has highlighted the need for clearer legal standards regarding synthetic evidence.
South Korea’s experience with deepfakes extends beyond this incident. Earlier in 2025, AI‑generated audio was used in a political context to spread false allegations against a chief justice, prompting calls for stricter regulation. In March 2026, the government announced plans to deploy advanced AI detection tools to counter deepfake video and audio crimes.
The Kim Soo‑hyun case underscores the broader challenge of verifying AI‑generated content in a fast‑moving information environment. While forensic experts can eventually detect manipulation, the lag between release and verification can allow false narratives to shape public perception and damage reputations.
As the legal system grapples with synthetic evidence, the entertainment industry and the public will likely demand greater transparency and accountability from content creators. The outcome of the ongoing investigation and any subsequent court rulings may set precedents for how defamation cases involving AI‑generated material are handled in South Korea.
The case remains unresolved, but it has already prompted discussions about updating defamation law, improving forensic capabilities, and instituting safeguards against the misuse of AI in public discourse.
The audio, posted on March 10, 2025, claimed that Kim Soo‑hyun and Kim Sae‑ron had dated from middle school to college and that their first sexual encounter occurred during a winter break in Kim Sae‑ron’s second year of middle school. The clip was used by Kim Se‑eui to accuse Kim Soo‑hyun of defamation. Within eight days, 2,026 news outlets reported on the allegation, prompting Kim Soo‑hyun to hold a press conference in which he denied the claims.
Police investigation revealed that the file submitted for forensic analysis was not the original recording. The National Forensic Service stated it could not determine whether the audio had been manipulated because the file was incomplete. However, investigators concluded that the recording was fabricated based on several factors: multiple conflicting versions of the clip circulated, the distribution trail suggested manipulation, and KakaoTalk messages linked to the case appeared altered.
Attorney Ko Sang‑rok, representing Kim Soo‑hyun, said investigators considered the existence of “completely different versions of a recording said to capture the same conversation on the same day and at the same time” distributed to multiple parties. Former prosecutor Lee Chang‑hyun noted that while proving fabrication in court is challenging, advances in analytical technology may eventually allow reliable detection of AI‑generated evidence.
Professor Yoo Hyun‑jae of Sogang University warned that the case demonstrates how AI‑generated material can be weaponised for defamation. He said the incident “has lifted the lid on Pandora’s box” and that without institutional safeguards, more individuals could use similar techniques to influence public opinion.
The case has attracted attention from law enforcement and the media. Kim Se‑eui was arrested and sent to prosecutors on charges of defamation and using AI‑manipulated audio. His real estate was reportedly locked by court order, and the case has highlighted the need for clearer legal standards regarding synthetic evidence.
South Korea’s experience with deepfakes extends beyond this incident. Earlier in 2025, AI‑generated audio was used in a political context to spread false allegations against a chief justice, prompting calls for stricter regulation. In March 2026, the government announced plans to deploy advanced AI detection tools to counter deepfake video and audio crimes.
The Kim Soo‑hyun case underscores the broader challenge of verifying AI‑generated content in a fast‑moving information environment. While forensic experts can eventually detect manipulation, the lag between release and verification can allow false narratives to shape public perception and damage reputations.
As the legal system grapples with synthetic evidence, the entertainment industry and the public will likely demand greater transparency and accountability from content creators. The outcome of the ongoing investigation and any subsequent court rulings may set precedents for how defamation cases involving AI‑generated material are handled in South Korea.
The case remains unresolved, but it has already prompted discussions about updating defamation law, improving forensic capabilities, and instituting safeguards against the misuse of AI in public discourse.
