South Korea's bold foray into regulating artificial intelligence has sparked a heated debate, with some hailing it as a model for the world and others criticizing its limitations. The country's newly enacted AI basic act, which took effect last week, is aimed at promoting industry growth while minimizing regulatory hurdles. However, local tech startups and civil society groups have expressed concerns that the law may be too restrictive or not stringent enough.
Under the new legislation, companies providing AI services will be forced to label clearly artificial outputs such as cartoons or artwork, while realistic deepfakes require visible labels. High-impact AI systems used for medical diagnosis, hiring, and loan approvals must also undergo risk assessments and document their decision-making processes. Extremely powerful AI models are required to submit safety reports, although the threshold is set so high that no current models worldwide meet it.
The government has promised a one-year grace period before penalties are imposed, with fines of up to $15,000 for non-compliance. However, critics argue that the process of self-determining whether systems qualify as high-impact AI creates uncertainty and competitive imbalance, as all Korean companies face regulation regardless of size.
Civil society groups have expressed concerns that the law does not provide sufficient protection for people harmed by AI systems, particularly in light of South Korea's alarming rates of deepfake pornography and exploitation. The country has accounted for 53% of global deepfake victims according to a recent report.
While some experts see the new legislation as a positive step towards promoting industry growth while ensuring safety and accountability, others argue that it falls short in several areas. Unlike other jurisdictions like the EU or US, which have adopted stricter risk-based regulatory models, South Korea has opted for a more flexible framework centered on "trust-based promotion and regulation".
The law's origins date back to 2020, but previous versions stalled due to provisions prioritizing industry interests over citizen protection. The current version has been criticized by civil society groups for lacking clear definitions of high-impact AI and providing limited protection for users.
As South Korea strives to become one of the world's leading AI powers alongside the US and China, it remains to be seen whether its new legislation will prove effective in balancing industry growth with public safety and accountability.
Under the new legislation, companies providing AI services will be forced to label clearly artificial outputs such as cartoons or artwork, while realistic deepfakes require visible labels. High-impact AI systems used for medical diagnosis, hiring, and loan approvals must also undergo risk assessments and document their decision-making processes. Extremely powerful AI models are required to submit safety reports, although the threshold is set so high that no current models worldwide meet it.
The government has promised a one-year grace period before penalties are imposed, with fines of up to $15,000 for non-compliance. However, critics argue that the process of self-determining whether systems qualify as high-impact AI creates uncertainty and competitive imbalance, as all Korean companies face regulation regardless of size.
Civil society groups have expressed concerns that the law does not provide sufficient protection for people harmed by AI systems, particularly in light of South Korea's alarming rates of deepfake pornography and exploitation. The country has accounted for 53% of global deepfake victims according to a recent report.
While some experts see the new legislation as a positive step towards promoting industry growth while ensuring safety and accountability, others argue that it falls short in several areas. Unlike other jurisdictions like the EU or US, which have adopted stricter risk-based regulatory models, South Korea has opted for a more flexible framework centered on "trust-based promotion and regulation".
The law's origins date back to 2020, but previous versions stalled due to provisions prioritizing industry interests over citizen protection. The current version has been criticized by civil society groups for lacking clear definitions of high-impact AI and providing limited protection for users.
As South Korea strives to become one of the world's leading AI powers alongside the US and China, it remains to be seen whether its new legislation will prove effective in balancing industry growth with public safety and accountability.
... on one hand, i get that they wanna promote industry growth but at what cost? 
it feels like the gov is still playing catch up on ai safety 
. the fact that deepfakes are super prevalent in south korea and now they're gonna label them with a warning sign? 
not gonna stop people from exploiting others, you know? 
also, what's with the "trust-based promotion" framework? does that even make sense? 
seems like it's just gonna let big corps off the hook while small startups get left behind... 
gotta keep an eye on this one, tbh 
. and fines of $15k seem pretty reasonable for a country that's still catching up on tech 
. maybe they should've asked the public more directly about what they want, but at least they're trying 
.
. And what about the threshold for reporting safety concerns? It's so high that it's basically saying "good luck" 
. I just hope that South Korea can figure out this AI thing before it gets completely out of control 
.
. They could definitely learn from each other's experiences...maybe the EU or US have better systems in place? 
... on one hand, we gotta protect ourselves from harm, but on the other hand, we can't suffocate innovation just because it's new & exciting 
... south korea's trying to find that middle ground with their ai regulation law, but it's not perfect, and that's what makes it relatable 
... maybe they're learning from their mistakes & trying to do better? 
only time will tell if this law really works in the long run 
. I'm not sure if their "trust-based promotion and regulation" framework will work as well as other countries' risk-based models. The problem with lack of clear definitions of high-impact AI is that it's hard to regulate something you can't even define properly 
.
, but what about the actual impact on society? Like, what if someone's life is ruined by a deepfake? The law seems to be more focused on protecting industry interests than actual people 
. I need some real sources behind these claims before I can even consider this a model for the world 
, but on the other hand, you got people getting hurt by deepfakes and stuff 
.
. And the law's all about promoting industry growth, but at what cost? 
and nobody knew what to do 
, especially since deepfakes have become a huge issue there. The labels on cartoons and artwork are gonna help consumers make informed decisions 
. The law needs better protection for them 
. It's not just about the tech companies; it's about the people who get hurt by these systems 
. This "trust-based promotion" framework doesn't quite sit right with me 
, but only time will tell if it's the right one 
.
... "trust-based promotion and regulation" sounds like a bunch of corporate speak for "we're not really doing anything about the risks". I mean, who doesn't want to be protected from deepfakes and AI-powered exploitation? But until they can even define what high-impact AI is, it's all just hot air. And $15k fines are basically peanuts compared to the potential damage of a faulty AI system. The EU and US are killing it with their risk-based regulatory models, Korea should take notes 
. But at the same time, I think the threshold for high-impact AI is kinda low 
. I mean, some companies might be able to game the system and get away with stuff without proper risk assessments.
. Maybe South Korea can learn from them and create an even better framework that balances industry growth with public safety 
. Plus, the fine of $15k is pretty paltry compared to how much some of these companies are making off this AI tech. And don't even get me started on the whole "trust-based promotion and regulation" thing - that sounds like just corporate-speak for "we're gonna let you play fast and loose as long as we're happy with it 
".
.
. The whole labeling thing is a good start, but what if some companies just find ways around it? And the lack of clear definitions on high-impact AI has me worried - how are we supposed to know when an AI system is being used in a way that's really harming people? 
. These safety reports and risk assessments need to be super detailed so we can trust that companies are being honest about their techs. And what's up with the $15k fine? Like, that's nothing compared to what people could lose if they get sued for deepfake-related issues