Who is Responsible When AI Gets the Facts Wrong?

A German court has ruled that Google can be held directly liable for false claims generated by its AI Overviews feature, in a decision that may have implications well beyond search.

The Munich Regional Court issued a temporary injunction after Google’s AI-generated summaries falsely linked two Munich-based publishers to alleged scams, subscription traps, and questionable business practices. As reported in The Decoder, the court found that those claims did not appear in the underlying linked sources but were instead generated by Google’s AI system.

Traditional search engines have often been treated as intermediaries, directing users towards information published elsewhere. AI summaries operate differently. They rewrite, structure, and synthesise information into new statements. This provided the grounds for the court to deem Google’s AI Overview its own content.

AI Summaries Are Not Search Results

The central argument in the ruling is that AI Overviews are not equivalent to conventional search results. Whereas a search engine provides links, an AI-generated summary produces a self-contained answer.

In this case, the court found that Google’s system had created claims that did not appear in the linked sources. The overview reportedly presented those claims in a structured format, including summaries and warning signs, making the output understandable on its own.

That distinction allowed the court to reject protections traditionally applied to search engines. Google was ordered to stop repeating the claims and was held responsible for 80% of the legal costs. The company has said it intends to appeal.

Why AI Accountability Extends Beyond Search

As AI systems become more widely used to summarise information and support decision-making, accountability is becoming harder to separate from functionality.

The same principle could apply to AI assistants, enterprise co-pilots, research tools, knowledge management systems, and conversational search platforms. Any system that synthesises sources and presents new explanatory content may face scrutiny over whether it is retrieving information or creating it.

That distinction may become increasingly important in enterprise environments, where AI-generated outputs are already being used to support research, analysis, and operational workflows.

Trust is Becoming a Governance Issue

The ruling also highlights a growing tension in how information systems are expected to operate.

Denas Grybauskas, Chief Governance and Strategy Officer at Oxylabs, argued that users are increasingly looking for answers rather than links.

“The modern user isn’t just looking for a link; they are looking for synthesis and answers,” said Grybauskas.

That expectation creates a higher burden on reliability. Grybauskas emphasised that moving forward, AI developers will need to prove that their systems are safe and verifiable.

“This means moving toward rigorous, controlled data curation that minimises hallucinations and mitigates the risk of propagating misinformation, as the legal consequences for AI speech are only going to grow more significant,” Grybauskas added.

What AI Accountability Means for Enterprises

One of the most notable aspects of the ruling is that the court rejected the idea that users should verify AI-generated claims themselves.

The judges argued that if users must independently validate every answer, the value of the AI summary becomes less clear.

Many organisations are deploying AI tools because they promise faster synthesis, search, and decision support. If every output requires full manual verification, the efficiency gains become more difficult to justify.

This does not mean AI systems must be perfect.

It does suggest that accuracy, traceability, and governance are moving from desirable product features towards operational requirements.

The Munich case may ultimately be remembered less for its impact on Google than for the principle it attempts to establish. If an AI system generates a statement that cannot be traced back to its sources, responsibility may sit with the organisation operating that system.