'But the AI Said It' — The Day That Defense Got Shredded in a German Courtroom
A Munich district court ruled on May 28, 2026 that Google's AI Overviews constitute the company's own original speech — not third-party content — making Google directly liable for six fabricated claims that falsely labeled two Munich publishers, Verlagshaus24 and GeraMond, as fraudulent businesses operating subscription traps and billing scams. The court rejected the application of traditional search engine immunity principles, finding that a system which evaluates disparate sources and generates "an independent, new, substantive statement" belongs to a fundamentally different legal category than a link aggregator, and therefore cannot shelter behind platform immunity doctrines built for passive conduits. Penalties under the ruling include fines of up to 250,000 euros per violation and up to two years in prison for executives — stakes that become staggering when applied to a platform serving 2.5 billion monthly users whose 9% error rate produces approximately 57 million inaccurate answers per hour. The ruling's core principle — if you built the AI, deployed it, and control its algorithm, you legally own its speech — applies with identical force to ChatGPT Search, Perplexity, Microsoft Copilot, and every other generative AI search product currently operating at scale. Just as the 1995 Stratton Oakmont v. Prodigy verdict unexpectedly created the Section 230 immunity framework that shaped 30 years of internet law, the Munich ruling appears positioned to trigger the development of an entirely new legal category for AI-generated content — one that sits between publisher and platform in ways 20th-century law was never designed to handle.