Who’s actually liable when an AI spits out nonsense — the company that deploys it, or no one? The question sounds academic until a court answers it. That’s just happened in Germany, and Simon Willison picked up Bruce Schneier’s fitting commentary on it.
What happened
A German ruling decided that Google’s AI Overviews — those AI-generated summaries at the top of search results — count as Google’s own statements. The consequence: Google is liable for errors in those answers. No “that was just the AI,” no hiding behind the algorithm. What’s up there is Google’s word.
Bruce Schneier puts the underlying principle crisply: an AI agent is an agent of the person or organization that deploys it — and should be treated that way by the law. If a company hired human writers for its summaries, it would be liable for their mistakes. Why should that be any different just because the summary comes from a machine?
The logic behind it
Schneier’s argument has an economic core that makes sense to me. Letting businesses hide behind “faulty AI” would be a massive handout — and a disastrous incentive. Why pay human writers, lawyers, or doctors if AI is not only cheaper but also absolves the employer of any responsibility? Liability doesn’t vanish just because a machine wrote the text. It has to stay with whoever sends the machine out into the world.
This is more than a search-engine footnote. We’re all building agents right now that write emails on their own, ship code, and tee up decisions. If responsibility for that stays squarely with the operator, it changes how carefully these systems get built and deployed.
My take
I find this clarity healthy. Not because I want to slow AI down — quite the opposite. But because responsibility and trust belong together. A tool nobody stands behind isn’t, over time, a tool you can trust. That a court is applying a simple, old rule here — you’re responsible for what you publish — reassures me more than it worries me. Anyone using AI seriously should stand behind what comes out of it anyway.
Sources: Simon Willison: AI and Liability, Bruce Schneier: AI and Liability