Experts say AI companies can be sued under current law, but rules are unclear

In short: People harmed by AI can often sue using existing legal claims, but there is no single clear rule that automatically makes AI companies responsible.

What's going on

Lawyers and scholars say many AI related harms can fit into familiar types of lawsuits. These include negligence (carelessness), product liability (a defective product), and defamation (harmful false statements). But the result often depends on the details, like what the company built, how it was used, and what the company knew or should have expected.

Applying these laws to AI can be hard. In negligence cases, a person usually has to show the company had a duty to act carefully, failed to do so, and caused harm. With AI, it can be unclear what “reasonable” safety steps look like and whether a specific AI output actually led to the injury.

Product liability can also be complicated. Courts do not always agree on whether software or an AI service counts as a “product” in the first place. Even if it does, proving a “defect” is tricky because AI systems can give different answers at different times, more like a tool that changes its behavior than a toaster that either works or does not.

There are also questions about who should be blamed. The developer, the company that deploys the system, the platform hosting it, and the user may all play a role.

What to watch

In the US, courts are starting to allow some AI lawsuits to move forward, including a 2025 ruling in Garcia v. Character that let a product liability claim tied to a chatbot proceed. In the EU, a revised Product Liability Directive will explicitly cover software, including AI, and countries must put it into local law by December 9, 2026. Even then, people will still need to prove key facts, like cause and responsibility, not just that “AI was involved.”

Source: NYTimes