The development of generative AI relies on large-scale corpora that often contain personal information, making anonymization central to data-protection compliance. In comparative law, anonymization generally determines whether personal-information protection rules apply, with the key inquiry turning on identifiability and the reasonable likelihood of re-identification. In generative AI, however, parameterized learning, dynamic content generation, and long-term interaction have transformed anonymization from a static processing outcome into a probabilistic and context-dependent risk condition. This creates three main challenges: a mismatch between formal anonymization and substantive risk, which may lead to identity disclosure or sensitive-attribute inference; the erosion of anonymization’s function as a regulatory boundary, which may enable regulatory circumvention and undermine public trust; and uncertainty over responsibility allocation, which may hinder risk prevention, enforcement, and remedies. Existing rules offer only limited responses to residual risks, jointly produced risks, and dynamic changes in model operation. Accordingly, anonymization should be reconceptualized as an ongoing governance framework based on context-sensitive identifiability standards, tiered legal consequences, responsibility allocation, and lifecycle oversight, thereby reconciling personal-information protection with AI development.

The article presents two parallel Polish models of the protection of life in the prenatal phase under criminal law. This differentiation is due to the permissibility of the in vitro fertilization (IVF) procedure and modern diagnostic possibilities. When fetal impairment is detected, the statutes under analysis provide for different consequences depending on the origin of the embryo. The article presents several ways of resolving the axiological disharmony, indicating the reinstatement of the possibility of the termination of pregnancy on the grounds of fetal impairment as the most appropriate.

This article argues that Australia can pursue auditable accountability for public-sector artificial intelligence without enacting a single comprehensive AI statute, provided that existing legal duties, policy frameworks, standards, and procurement mechanisms are organised into an explicit assurance stack. An “AI Act” is used here to mean a unified statute that imposes system-wide ex ante obligations on providers and deployers. Australia currently relies on a more distributed model. The article’s claim is that this model can still be made defensible, but only if abstract norms are translated into evidence disciplines that preserve contestability and reviewability. Using a doctrinal and functional method, the article shows how legality, procedural fairness, record-making, reason-giving, procurement discipline, privacy obligations, and audit practices can be aligned to produce a minimum reviewable trace for AI-influenced public decisions. The original contribution is twofold. First, the article conceptualises Australia’s public-sector AI governance arrangements as an assurance stack whose layers only matter if they generate reviewable artefacts. Second, it proposes a bounded minimum reviewable trace that preserves the system configuration, material inputs and outputs, evaluation basis, reliance statement, and contestability pathway for a particular decision. The article also reframes selective hardening as a practical governance response for higher-risk uses, and presents an Assurance Requirement Level as a qualitative policy heuristic rather than a quantitative model. Concrete illustrations drawn from welfare eligibility and emergency-care triage demonstrate how the trace would work in practice. The article concludes that procurement is the principal operational lever for pushing evidence duties upstream, but that the stack will only succeed if audit offices, tribunals, ombudsmen, and agencies are equipped to interpret and test the artefacts they require.