Private Law Theory

ABSTRACT
Legal responses to failures of generative artificial intelligence have focused primarily on the professionals who use AI-generated material. Lawyers have been sanctioned for fabricated authorities, firms have faced malpractice exposure, and courts have emphasized duties of competence, candour, and verification. Far less attention has been paid to the independent liability of the companies that design, market, integrate, and sell professional legal AI. This article examines the potential producer liability of legal AI manufacturers under the common-law systems of the United States, Canada, and England. Drawing upon product liability, negligence, misrepresentation, failure-to-warn principles, contractual defences, and selected emerging disputes, it rejects the assumption that professional responsibility automatically displaces manufacturer responsibility. Liability may be cumulative: the professional may breach a duty to verify while the producer simultaneously breaches duties arising from superior technical knowledge, product representations, commercial benefit, and architectural control. General disclaimers cannot necessarily neutralize foreseeable or inherent defects affecting a product’s intended professional use. The article applies the published Instrumentum Vocale doctrine to legal AI. An LLM is neither a legal actor nor an epistemic agent, but a speaking instrument operating within a human and corporate chain of responsibility. The article therefore proposes an external architecture based on assertion classification, an epistemic seam between generation and verification, qualified human review, source-level validation, provenance, and reconstructible audit trails. The decisive issue is not whether the AI can be blamed, but whether those who manufactured and controlled the instrument may escape responsibility by attributing its failures exclusively to downstream users.

Kildeev, Adel, The Tool Has a Manufacturer: Producer Liability for Legal AI (August 31, 2026).