Kim Soo-hyun case warns Thailand: AI deepfakes can ruin lives before the law catches up

What began as another South Korean celebrity scandal has become a warning for every country, including Thailand: in the age of artificial intelligence, online defamation no longer depends only on rumours. It can now arrive dressed as “evidence” — a familiar voice, a leaked chat, a screenshot, a clip that looks convincing enough to turn public opinion before investigators have time to test whether it is real.

The case centres on Kim Se-ui, operator of the conservative YouTube channel HoverLab, also known as the Garo Sero Research Institute, who has been detained in South Korea over allegations linked to actor Kim Soo-hyun. A Seoul court issued an arrest warrant in late May, citing concerns that the suspect could flee or destroy evidence.

From celebrity scandal to AI-evidence crisis

South Korean police say Kim Se-ui spread allegations through YouTube and other platforms claiming that Kim Soo-hyun had been in a relationship with the late actress Kim Sae-ron when she was a minor, and that pressure over debt was connected to her death. Police also accused him of using artificial intelligence to manipulate the late actress’s voice in material presented as evidence. Kim Se-ui has denied the allegations.

The case moved further when Seoul’s Gangnam Police Station referred Kim Se-ui to prosecutors while in custody. Korean reports say the charges include defamation, intimidation, attempted coercion and offences related to the distribution of unlawful material.

Crucially, this is not yet a final court verdict. The arrest and referral to prosecutors are legal steps in the criminal process, not a conviction. But the case has already raised a far larger question: what happens when AI-generated material is used not simply to mislead the public, but to frame a person in the court of public opinion?

Fake evidence, real damage

According to The Korea Times, police confirmed that the allegations relied on fabricated evidence, including doctored chat logs and an AI-generated voice recording. That finding changes the nature of the scandal. This is no longer only about whether a public figure was defamed. It is about whether digital tools can now manufacture a false trail convincing enough to damage a career, reputation and mental wellbeing before the truth catches up.

Kim Soo-hyun had already faced severe commercial fallout. Reuters reported in March 2025 that brands including Prada and Dinto cut ties with the actor amid the controversy, while his agency denied claims that he had dated Kim Sae-ron when she was underage.

The financial stakes are also significant. The Korea Times reported that the actor’s damages claim was expected to expand from 12 billion won to around 30 billion won, with advertisements, overseas schedules and content projects affected. The same report noted disruption to Knock-Off, a Disney+ series with a reported 60-billion-won budget.

The Thai warning

Thailand should not read this only as Korean entertainment news. The same pattern could hit a Thai actor, politician, executive, influencer, journalist, teacher, student or ordinary citizen. A fake voice note, edited Line chat, manipulated image or AI-generated video could spread in minutes. The victim may spend months proving something that the public believed in seconds.

This is the dangerous gap: technology moves at platform speed, while justice moves at institutional speed.

Thailand already has legal tools that could be relevant. Criminal defamation under Sections 326 and 328 of the Thai Criminal Code covers statements that damage reputation, with higher punishment when defamation is committed through publication or wider distribution. The Computer Crime Act also addresses forged or false computer data, electronically modified images that cause reputational harm, and the use of computer data as evidence under proper legal conditions.

But the Kim Soo-hyun case shows that punishment after the fact may not be enough. The harder question is speed: can Thai investigators authenticate AI-generated audio, video and screenshots quickly enough to prevent irreversible reputational harm?

Laws exist, but verification is the battlefield

Thailand has also begun building AI-governance infrastructure. ETDA’s AI Governance Center says its mission includes developing AI governance suited to Thailand’s context and aligned with international policy, while also providing consultation and knowledge-sharing on responsible AI use. ETDA has also promoted guidelines for the responsible organisational use of generative AI, covering benefits, limitations, risks and governance.

That is a useful foundation. But deepfake defamation requires more than broad AI ethics. It requires operational readiness: forensic tools, trained officers, fast evidence-preservation procedures, cooperation with platforms, and clear standards for how courts should treat audio, images and chats that may have been synthetically generated.

Internationally, regulators are moving in the same direction. The EU’s AI Act transparency obligations under Article 50 require marking, detection and labelling of AI-generated content and deepfakes, with those obligations applying from August 2, 2026. The lesson for Thailand is not to copy every foreign rule, but to recognise the same core problem: synthetic media must be identifiable before it becomes social truth.

The real issue: evidence can now be manufactured

The Kim Soo-hyun case captures a new phase of online harm. In the past, defamation often meant an accusation, a rumour or a distorted narrative. Now, AI can add the missing “proof”: a voice that sounds real, a chat that looks familiar, a photo that appears difficult to deny.

That changes the burden on victims. It is no longer enough to say, “That is false.” They must prove that the evidence itself was manufactured — often while the internet has already reached its verdict.

For Thailand, the urgent question is not only whether existing law can punish a deepfake offender. It is whether Thai institutions can respond before a fake clip becomes a national scandal, before sponsors withdraw, before employers panic, before families are harassed, and before a person’s name is permanently damaged online.

The Kim Soo-hyun case leaves one clear warning: in the AI era, justice must not only be fair. It must be fast enough to matter.