Comparative Law Was Built for Slow Systems. AI Isn't One.

The German Law Journal published Volume 26, Special Issue 7 - "Comparative AI Law: Regulating the Future" - in October 2025, and Cambridge Core flagged it again on May 1, 2026. Fifteen contributions. Authors from Beijing to Brussels, Silicon Valley to Sydney. Editor Emanuel V. Towfigh, splitting time across EBS Universität, Peking University's School of Transnational Law, and the Max Planck Institute.

The premise breaks with the dominant frame. Towfigh argues the "AI race" narrative - whoever regulates first wins, whoever deregulates fastest dominates - misreads what governments are actually doing. They're not racing. They're running parallel experiments on technology that ignores borders and traditional legal categories.

That reframing matters because it shifts what comparative scholarship is for. Not picking a winner. Reducing the cost of learning from each other's failures fast enough to keep up with systems evolving by the microsecond.

What's actually in the issue

The volume lands across four pressure points. The EU AI Act's social scoring ban under Article 5(1)(c) - widely sold as a rebuke to China's Social Credit System, but drafted vaguely enough to potentially capture ordinary credit scoring and insurance underwriting. European Commission guidelines from early 2025 already lean toward a narrow reading. One contribution argues the EU should drop the dystopian framing and learn from China's own SCS regulatory experience instead.

Copyright comes in through the Beijing Internet Court's 2023 ruling - the world's first judicial decision that AI-generated content can qualify for copyright if a human contributed demonstrable intellectual input. Compared against the US Copyright Office's Zarya of the Dawn refusal. Same technology. Opposite outcomes.

Anti-discrimination law gets a harder treatment. Generative AI doesn't just amplify the bias problems algorithmic decision systems already had. It introduces new categories of harm that current frameworks weren't built to recognize. The AI Act's heavy reliance on self-regulation gets criticized directly - it works only when backed by credible enforcement, and regulators in most jurisdictions can't match the technical depth of the firms they're supposed to oversee.

Governance, not just legislation

A separate piece on AI governance argues existing oversight arrangements in the legal profession aren't fit for purpose. Confidentiality and accuracy issues are too significant to delegate to whatever guardrails a vendor ships with their product. The article sketches what good organizational AI governance actually looks like, cross-referencing Canada's Algorithmic Impact Assessment Tool, the NSW AI Assessment Framework, South Korea's AI Framework Act, Japan's 2025 bill on AI promotion and risk, and the UK's pro-innovation white paper.

It's a reminder that legislation is the visible layer. Most of what determines whether AI causes harm sits one level down - in procurement standards, impact assessments, internal codes, and supervisory mechanisms that don't make headlines.

What the comparative lens exposes

The cross-jurisdictional reading does something individual national debates can't. It surfaces the fact that everyone is improvising with the same gaps: liability allocation across the AI value chain, training-data accountability, redress for algorithmic discrimination, IP rights in synthetic outputs. The EU went prescriptive. The US went sectoral. China went directive-led. The UK went principles-based. None has solved any of those problems.

What the issue argues, through fifteen different lenses, is that these aren't competing answers. They're fragments of an answer no single jurisdiction has yet assembled.

The unresolved part

The harder question the issue circles without resolving: comparative law works when systems are slow enough to compare. AI systems aren't. By the time a regulatory sandbox in one jurisdiction produces useful evidence, the underlying model has been deprecated twice over. The pacing problem isn't a bug in any specific regulatory approach. It's the condition all of them are operating under.

So the real test isn't whether Brussels learns from Beijing, or Washington from Sydney. It's whether any of them can learn fast enough at all.

Sources: Cambridge Core Blog (May 1, 2026); German Law Journal Vol. 26, Special Issue 7, "Comparative AI Law: Regulating the Future" (October 2025); Towfigh, "Comparative AI Law: An Introduction"; "AI Governance: A Comparative Approach"; "Regulatory Alternatives to the AI Social Scoring Ban: A Comparative Perspective"; "Digital Alchemy? Rethinking Copyright in the Age of AI-Generated Content."