W.J. Jon (additional co-authorship, possibly including a “D. Kim” as in the original note, unconfirmed — I could verify WooJung Jon as an author via a researcher profile page, but could not independently confirm a “D. Kim” co-author from the sources I accessed) — full title: “How ‘hard’ are hard laws? AI legislation, soft-law governance, and comparative lessons from South Korea and Japan,” Computer Law & Security Review (ScienceDirect, 2026). Venue, full title, and WooJung Jon’s authorship confirmed via the ScienceDirect article listing (DOI path S2212473X26001069) and Jon’s academic profile (Centre for Digital Law); the ScienceDirect page itself returned a 403 to direct fetch, so I was working from search-indexed summaries rather than the full text — flagging this as a partial-access verification.

Core Contribution

The title states the thesis as a question, and that’s the analytical move: South Korea’s AI Basic Act (passed December 2024, the first comprehensive AI statute enacted outside the EU) is often described as “hard law” — binding, enforceable regulation — in contrast to Japan’s historically soft-law, guideline-driven approach. This paper interrogates whether that hard/soft distinction actually predicts real-world enforceability, using South Korea and Japan as a paired comparative case study. The Korean AI Basic Act itself is structurally hybrid — it combines soft-law, ethics-oriented provisions (its Articles 27–30) with hard-law regulatory obligations (Articles 31–36) inside a single statute — which is itself evidence for the paper’s likely core claim: the hard/soft binary is too coarse, and legal “hardness” is a matter of institutional implementation capacity as much as statutory text.

Method

Comparative legal and policy analysis — doctrinal review of the statutory text and enforcement architecture in both jurisdictions (South Korea’s AI Basic Act; Japan’s soft-law guidelines and its more recent floated Basic Act on the Advancement of Responsible AI, which would mark a Japanese shift toward hard law if pursued). This is not an empirical/quantitative study of enforcement outcomes — no dataset of enforcement actions, penalties levied, or compliance rates that I could confirm — it’s a legal-institutional argument about the design and likely functional effect of the statutory architecture itself, including whether Korea’s hybrid hard/soft structure functions as intended or ends up defaulting to its softer provisions in practice.

Limitations

  • Purely legal/doctrinal framing — no technical engagement with whether the policies discussed are mechanically enforceable at the model or data layer (e.g., no discussion of audit tooling, technical verification standards, or compliance-checking infrastructure that I could confirm).
  • Two-country comparison (South Korea, Japan) — generalization to other jurisdictions’ hard/soft law debates (EU AI Act, US state-level approaches) is not the paper’s stated scope.
  • I was unable to access the full text directly (ScienceDirect blocked automated fetch), so specific findings beyond the thesis and institutional description above are not confirmed — treat the “core contribution” section as the confirmed thesis, not a confirmed set of findings.
  • The full author list is only partially confirmed (WooJung Jon confirmed; the “D. Kim” co-author from the original note is unconfirmed by me).

Relevance to My Niche

Background/adjacent rather than core. This sits in the governance/policy layer of the site rather than the red-teaming/quantization/multilingual ML niche directly — it’s useful for thinking through how durable an AI-governance obligation (e.g., a data-attribution or worker-protection right) would need to be, structurally, to survive the same hard/soft enforceability critique this paper applies to Korea and Japan’s AI statutes. If any future policy argument is made about mandating red-teaming disclosure, quantization safety testing, or multilingual safety-evaluation requirements as law rather than voluntary industry practice, this paper is a relevant comparative case study for why the mere existence of a “hard law” provision doesn’t guarantee real enforceability — the institutional capacity behind it matters just as much.