When AI Causes Harm: Who’s Liable for an Injury?
Artificial intelligence now drives cars, screens medical images, and controls machinery on factory floors, and when those systems fail, real people get hurt. The technology is new, but the legal question is familiar: when AI causes harm, who is responsible?
When an AI system injures someone, liability usually falls on the humans and companies behind it: the manufacturer, the software developer, the business that deployed it, or an operator who misused it. Utah courts apply established negligence and product liability principles to these claims, so a novel technology does not erase a victim’s right to recover.
How AI actually causes harm
AI rarely fails in dramatic, science-fiction ways. More often it fails quietly: a self-driving feature that misreads a stopped vehicle, a diagnostic tool that misses a tumor, or an industrial robot that moves when it shouldn’t. Because these systems make decisions that used to belong to people, a single flaw can injure many before anyone notices.
- Automated vehicles. Driver-assist and self-driving features can misjudge hazards and cause crashes.
- Medical AI. Diagnostic and treatment tools can produce dangerous errors if they’re flawed or misused.
- Robotics and automation. AI-controlled equipment can injure workers and bystanders when safeguards fail.
Who can be held responsible
The key insight is that AI isn’t a legal person. It can’t be sued, and it can’t pay damages. Responsibility instead attaches to the people and organizations that designed, built, sold, or deployed it. Depending on the facts, more than one party may share the blame.
| Scenario | What went wrong | Who may be liable |
|---|---|---|
| Self-driving crash | System misread the road | Automaker, software developer, sometimes the operator |
| Faulty medical AI | Wrong or missed diagnosis | Device maker, software firm, provider who relied on it |
| Industrial robot injury | Safeguard or sensor failed | Manufacturer, integrator, employer |
| Defective algorithm | Flawed design or training data | Developer and the company that deployed it |
Key point. Companies sometimes hide behind the complexity of their technology, suggesting no one could have foreseen the failure. But the legal standard still asks whether they acted reasonably and whether their product was unreasonably dangerous, questions a thorough investigation can answer.
The legal theories behind an AI injury claim
Most AI injury cases rest on two familiar foundations. Product liability applies when a system is defectively designed, manufactured, or sold without adequate warnings, and it can hold makers responsible even without proof of carelessness. Negligence applies when a company or operator failed to act with reasonable care, for example, by deploying an undertested system or ignoring known risks. A single incident can support more than one theory at once.
How Utah law treats new technology
Utah has not thrown out its injury laws because a computer was involved. The same principles that govern defective cars and unsafe workplaces apply to AI-driven products, and the general deadline to file a personal injury lawsuit is typically four years from the date of injury, though shorter deadlines can apply in certain situations. Because these cases are complex and evidence can disappear, it’s wise to act early rather than wait.
What to do if an AI system injured you
Get medical care and document it
Prompt treatment protects your health and records the injury and its cause.
Preserve the evidence
Save devices, software versions, error logs, manuals, and anything showing what the system did.
Identify everyone involved
The maker, developer, and deploying company may all bear responsibility.
Talk to a lawyer quickly
An attorney can secure technical evidence before it’s overwritten or lost.