For years, technology companies have relied on a familiar argument: we are not responsible for the information we surface because we did not create it. And a recent court ruling in Germany may have just challenged that assumption.
Google’s AI just hit a legal turning point
The Munich Regional Court preliminarily ruled that Google can be held liable for false statements generated by its AI Overviews feature. The case involved two publishers who discovered that Google’s AI-generated summaries associated them with scams, questionable business practices, and subscription fraud despite that those claims did not appear in the underlying sources cited by Google. The court found that the AI had effectively created new statements by misinterpreting information gathered from different sources.
At first glance, this may seem like another legal dispute about technology, but in reality, it raises a much bigger question: What happens when AI is no longer simply retrieving information, but actively rewriting, reorganizing, and generating it?
A distinction that changes everything
For decades, search engines operated like librarians: they helped users locate information that already existed elsewhere, and if a website published something false, responsibility generally rested with the publisher who created the content.
According to the Munich court, AI Overviews operate differently. Rather than merely displaying links, the system creates “independent, new, and substantial statements” based on its interpretation of available information. In other words, it becomes an author of sorts, not just an index. This is the part of the ruling that I find most significant.
Generative AI companies have often framed their systems as tools that summarize existing information. Yet anyone who has spent enough time with large language models knows that summarization is only part of the story, as these systems routinely infer, synthesize, compress context, and sometimes fill gaps with information that never existed in the source material. The industry has a name for this phenomenon: hallucination. A word that sounds almost playful. It becomes a lot less playful when a business is falsely linked to fraud, when a professional’s reputation is damaged, or when misinformation spreads at internet scale.
Google reportedly argued that users are warned that AI-generated results may contain errors and should be independently verified. The court rejected that defense, noting that the disputed statements did not appear in the cited sources and that responsibility for correcting such misinformation could not reasonably fall on third parties.
When “verify it yourself” stops being a valid answer
Personally, I believe the court reached the right conclusion. Disclaimers are important, but they cannot become a universal shield against accountability. Imagine a newspaper publishing false accusations and then adding a footer that says, “Please verify the facts yourself.” Few people would consider that sufficient. The same logic should apply when information is generated by algorithms rather than journalists.
This does not mean AI companies should be punished for every mistake their models make. Perfection is neither realistic nor possible. It does mean that organizations deploying AI at scale should carry meaningful responsibility for the outputs they distribute, especially when those outputs are presented as authoritative answers. That responsibility becomes even more important because of how people interact with AI.
Research consistently shows that users tend to trust concise answers. Many never click through to source material. AI Overviews were designed precisely to satisfy that behavior by delivering immediate answers at the top of search results. When users receive information in that format, they are not treating it as a rough draft. They are treating it as an answer. And that is where the conversation extends far beyond Google.
If the legal reasoning behind this ruling gains traction internationally, every company building generative AI systems may face tougher questions about accountability.
OpenAI, Anthropic, Perplexity, Microsoft, and countless startups operate on a similar premise. They generate responses by synthesizing information from vast collections of data. Most also include warnings that outputs may be inaccurate. The German court’s reasoning suggests that warnings alone may not be enough when a system creates new claims that cause harm.
From capability to accountability
In many ways, this case represents a turning point in the evolution of AI governance. For the last few years, much of the discussion around artificial intelligence has focused on capability. Models became faster, larger, and more sophisticated. The dominant question was what AI could do, now the question is becoming what AI should be responsible for. And that shift is healthy.
Innovation without accountability eventually erodes trust. And trust is the foundation on which every AI product depends. As someone who spends a great deal of time thinking about content, knowledge, and information systems, I see this ruling as part of a broader transition. We are moving from an era where AI is treated as an experimental assistant to one where it increasingly functions as a public information infrastructure.
When technology reaches that level of influence, responsibility can no longer be optional.
The real test for AI
The Munich court may not have issued the final word on AI liability. Google has indicated that it is reviewing the decision, and the case could still evolve through appeals. Yet the significance of this ruling extends far beyond the outcome of a single lawsuit.
The important question is not whether AI can make mistakes (because we have known for years that it can), but who bears responsibility when those mistakes create tangible harm. As AI systems become embedded in the way people search for information, make decisions, and form opinions, the long-standing distinction between technology providers and content publishers is beginning to blur.
For much of the AI era, innovation has outpaced accountability. Companies raced to build more powerful models, while society largely focused on what these systems were capable of doing. That conversation is now evolving. Increasingly, the question is not what AI can do, but what its creators should be accountable for when it does it. Whether this ruling ultimately stands or not, it reflects a broader shift in public expectations. As AI moves from experimental assistant to trusted information intermediary, responsibility can no longer be treated as a secondary consideration. The future of artificial intelligence will be shaped not only by the intelligence of its systems, but by the willingness of the organizations behind them to stand behind their outputs.
Trust, after all, is not generated by algorithms, it is earned through responsibility.
