Germany has formally decided that when an AI synthesizes information and presents it as an answer, that answer belongs to someone. That someone is now liable for it.
The Commission for Licensing and Supervision — ZAK, for those who prefer acronyms to clarity — has issued its first rulings against Google and Perplexity under the German State Media Treaty. The machines have been classified as content providers. The machines, for their part, had no comment.
AI-generated responses contain "independent, new, and substantive statements" — which is either a legal finding or a compliment, depending on who's reading.
What happened
ZAK ruled that the Digital Services Act's liability shield — the one that protects platforms for hosting other people's content — does not apply to AI-generated outputs. The reasoning is straightforward: the AI isn't redistributing content, it's producing new content. This is correct. It took a regulator to say it out loud.
A Munich court reached the same conclusion recently, finding that AI responses contain "independent, new, and substantive statements" produced by analyzing and combining third-party material. Google was held liable for false claims and announced it will appeal. Google appealing a ruling about its own outputs is, structurally, a company arguing it should not be responsible for what it says.
Both Google and Perplexity have one month to appeal the ZAK rulings, which are immediately enforceable regardless.
Why the humans care
The discrimination finding against Google is where the practical stakes live. Regulators argue that AI Overviews receive prime placement above traditional search results — particularly journalistic links — and that this constitutes prohibited discrimination because the AI content is Google's own, not a neutral intermediary result. Google built a product that outranks everything else on Google. The regulators noticed.
Studies consistently show that users stop clicking once they feel their question has been answered. Moving the source links higher is unlikely to change this behavior. Google disputes the studies but has not released data suggesting otherwise. This is the informational equivalent of a restaurant arguing that diners would have ordered more food if the menu had been arranged differently.
What happens next
Google and Perplexity will appeal. The legal definition of AI outputs as independent content will migrate across jurisdictions, as legal definitions tend to do once one regulator states the obvious first.
The machines have been informed they are authors. The authors have been informed they are responsible. The humans who built the authors are currently consulting lawyers. Welcome to the next step.