A defect rooted in training data can be a design: Considering product liability law, which statement is most

A training-data defect can be a design or manufacturing defect, and inadequate warnings or failure to remediate raise the manufacturer's product-liability exposure.

The question

Preparing for AIGP? Take the free 5-min readiness quiz →

1. Liability is impossible because AI behavior is probabilistic, so any harm from an unpredictable model output is legally treated as user misuse rather than a product defect.

   Plausible-sounding but probabilistic behavior does not immunize a product; a design or manufacturing defect causing foreseeable harm can still ground product liability.
2. A defect rooted in training data can be a design or manufacturing defect, and inadequate warnings plus a failure to remediate through updates can each increase the manufacturer's exposure. ✓

   Correct: product liability reaches design/manufacturing defects and inadequate warnings, and an AI defect from training data with unaddressed post-sale risks heightens exposure.
3. Because the company issues post-sale software updates, it is fully shielded from liability for earlier harms, since the availability of a later patch retroactively cures any original defect.

   Almost plausible but a later patch does not retroactively cure harm already caused, and failing to remediate known risks can itself add exposure.
4. Responsibility rests solely with the component supplier that provided the training data, so the manufacturer that designed and sold the finished robot bears none of the product-liability risk.

   Plausible but the manufacturer that designs and places the finished product on the market generally bears product-liability responsibility, regardless of a supplier's role.

How to remember it

A training-data defect can be a design or manufacturing defect, and inadequate warnings or failure to remediate raise the manufacturer's product-liability exposure.

More Understanding How Laws, Standards and Frameworks Apply to AI questions

• Substantiate or drop the lowest-price claim: Under consumer protection law, which governance step is most →
• The emotion tool is prohibited in the workplace: Under the EU AI Act tiers, which classification is most →
• Operate a risk-management process and appropriate data: Which sequence best reflects the EU AI Act's →
• All 394 Understanding How Laws, Standards and Frameworks Apply to AI questions →