Compliance

Product Liability (AI)

Legal responsibility of manufacturers and suppliers when AI products or components cause harm due to defects.

Definition

Product Liability (AI) refers to the application of product liability law to AI systems, holding manufacturers, developers, and suppliers responsible when AI products cause harm due to defects. The EU has modernized its framework to explicitly address AI.

EU Product Liability Directive 2024/2853: Key updates for AI:

  • Software as product: Explicitly includes software, including AI systems
  • Defect definition expanded: Includes failure to meet safety requirements, lack of updates, and cybersecurity vulnerabilities
  • Learning systems: Manufacturers liable for defects arising from machine learning after deployment
  • Disclosure: Courts can order manufacturers to disclose evidence

Types of AI Product Defects:

  • Design defects: Flawed algorithm architecture or training approach
  • Manufacturing defects: Errors in implementation, data corruption
  • Warning defects: Inadequate disclosure of limitations or risks
  • Update failures: Not providing necessary safety updates

Liable Parties:

  • AI system manufacturer/developer
  • Hardware manufacturer (if relevant)
  • Importer (for non-EU products)
  • Supplier (if manufacturer cannot be identified)

Relationship with AI Act: Non-compliance with EU AI Act requirements may constitute evidence of defect under product liability law.

Related concepts: AI Liability, Strict Liability, Provider, Deployer

Sources

  • EU AI Liability Directive
  • Product Liability Directive
  • 85/374/EEC