A German court treats Google's AI as its agent

Simon Willison flags Bruce Schneier on a recent German ruling that Google can be held liable for errors its AI overviews introduce. Schneier’s framing is the part worth keeping: “AI agents are agents of the person or organization that deploys them — and should be treated by the law as such.” If a company hired human writers to produce its summaries, it would own their inaccuracies; letting the same company hide behind faulty AI, he argues, would be a handout and an invitation to misbehave.

The reason this lands on the trust beat is that “agent of whoever deploys it” is exactly the question the identity world is circling from the other direction — agent identity, delegated credentials, who-vouches-for-this-actor. Liability law getting there first, by simply mapping AI agents onto the old doctrine of agency rather than inventing a responsibility vacuum, is a useful anchor point to track as the credentialing side catches up.

It’s a single national ruling, not settled doctrine, but the principle is clean enough to be sticky — and worth watching whether US regulators and courts reach for the same agency framing.