AI Decision Liability Framework LiabilityChain Deep Dive: Who Bears Responsibility When AI Makes Wrong Decisions

The Accountability Vacuum

When an autonomous vehicle injures a pedestrian, who is responsible? When an AI medical diagnostic system misses cancer, who bears liability? As AI systems proliferate in high-risk scenarios, "who's responsible when AI errs" has become one of the most pressing issues facing global legal systems.

The UN Commission on International Trade Law (UNCITRAL) today released AI decision liability framework LiabilityChain, establishing a three-tier liability allocation mechanism.

Core Framework

Tier 1: Strict liability for AI deployers (companies operating autonomous fleets, hospitals deploying AI diagnostics) — no fault proof required, only causation between damage and AI decision.

Tier 2: Fault liability for AI developers and training data providers — liable when insufficient safety testing, biased/incomplete training data, failure to patch known vulnerabilities, or inadequate risk disclosure to deployers.

Tier 3: Right of recourse — deployers bearing strict liability may seek indemnification from at-fault developers or data providers, proportioned by fault severity.

Liability Proportion Determination

LiabilityChain's most innovative element is "AI decision traceability" requirements. All high-risk AI systems must maintain decision logs recording input data, reasoning paths, and output results for each decision. When damages occur, decision logs serve as key evidence for determining liability proportions.

Global Legislative Race

LiabilityChain accelerates global AI liability legislation. The EU has incorporated similar provisions into its AI Act effective 2030. The US is discussing the AI Liability and Transparency Act. China is developing its AI Law with tort liability provisions.