Munich, Germany — In a ruling that reverberates across the tech and creative industries, a Munich Regional Court has declared OpenAI liable for copyright infringement after using protected song lyrics to train its ChatGPT models without authorization. The decision, handed down on Tuesday, November 11, 2025, marks the first major European court victory against the U.S.-based AI giant in a case centered on generative AI's use of artistic works. The court ordered OpenAI to pay undisclosed damages to GEMA, Germany's preeminent music rights organization, emphasizing that both the "memorization" of lyrics in AI training data and their reproduction in chatbot outputs violate German copyright law.
The case, filed by GEMA in November 2024, accuses OpenAI of systematically scraping and utilizing lyrics from its vast repertoire to fuel ChatGPT's language models. GEMA, which safeguards the rights of over 100,000 composers, songwriters, and music publishers, represented the creators of nine emblematic German songs in the lawsuit. These included Herbert Grönemeyer's iconic hits "Männer" and "Bochum," Rolf Zuckowski's holiday staple "In der Weihnachtsbäckerei," and Helene Fischer's chart-topper "Atemlos" (originally by Kristina Bach). Presiding Judge Elke Schwager, in a detailed oral verdict lasting over an hour, rejected OpenAI's defenses outright, stating that the company's practices amounted to unauthorized reproduction and exploitation of protected content. "The internet is not a self-service store, and human creative achievements are not free templates," echoed GEMA CEO Tobias Holzmüller in a post-ruling statement, underscoring the decision's philosophical weight.
This verdict stems from GEMA's broader campaign to hold AI developers accountable for what it terms "intellectual property theft" in the era of machine learning. Founded in 1903, GEMA has long been a bulwark for German musicians, collecting royalties from streams, broadcasts, and performances worldwide. In recent years, as AI tools like ChatGPT exploded in popularity—boasting over 200 million weekly users by mid-2025—the organization grew alarmed at the unchecked ingestion of copyrighted materials. GEMA's lawsuit highlighted how simple user prompts could elicit verbatim reproductions of lyrics, such as entire verses from Grönemeyer's anthems, demonstrating that ChatGPT had not merely "learned patterns" but retained and regurgitated specific protected texts.
OpenAI, valued at $157 billion and projecting $5 billion in 2025 revenue, mounted a vigorous defense. The San Francisco-based firm, co-founded by Sam Altman and initially backed by Elon Musk, argued that its large language models (LLMs) do not "store or copy" individual data points. Instead, they argued, ChatGPT synthesizes responses from probabilistic patterns derived from vast, anonymized training datasets—estimated at trillions of tokens scraped from the public internet. Regarding outputs, OpenAI contended that users, not the company, bear responsibility as the "producers" of generated content via their prompts. The court, however, found these claims unpersuasive, ruling that the EU's text and data mining (TDM) exception—intended for non-commercial research—does not shield commercial entities like OpenAI from licensing obligations. Judge Schwager expressed astonishment at the company's failure to anticipate the "clear legal situation" under German Urheberrecht (copyright law), which harmonizes with EU directives like the 2019 Copyright in the Digital Single Market.
The implications of this ruling extend far beyond the nine songs in question. GEMA's legal advisor, Dr. Kai Welp, described it as a "milestone" with "vital implications for the remuneration of creative artists" across Europe. By affirming that AI training constitutes a remunerable "reproduction right" and that outputs can infringe independently, the decision challenges the foundational business model of generative AI firms reliant on unlicensed data. It aligns with the EU AI Act, effective August 2025, which mandates transparency in training data for high-risk systems but stops short of explicit copyright mandates—leaving courts to fill the gap. Legal experts, including those from the Berlin firm Raue that represented GEMA, hail it as a "precedent-setting signal to the global tech industry," potentially influencing harmonized copyright enforcement from France to Finland.
This European milestone contrasts sharply with the patchwork of U.S. litigation, where OpenAI faces over a dozen suits but has notched partial wins. In June 2025, a New York federal court dismissed parts of a class-action by authors like John Grisham, citing "fair use" for transformative AI outputs, though training data claims persist. Media giants, including The New York Times and Thomson Reuters, continue to press cases alleging unauthorized scraping of news articles. Authors George R.R. Martin and Jonathan Franzen have joined similar actions, claiming ChatGPT "hallucinates" plagiarized passages. Bollywood labels in India and book publishers in Canada have filed analogous claims in 2025, signaling a global backlash.
In music specifically, the ruling echoes U.S. settlements like Universal Music Group's June 2025 accord with AI firm Udio, which agreed to undisclosed royalties for training on recordings. GEMA itself has a second lawsuit pending against Suno AI, a U.S. music generator, with a hearing set for January 26, 2026. Meanwhile, in September 2024, GEMA pioneered an AI-specific licensing model, offering tiered fees based on usage scale—training data access for €0.01 per 1,000 tokens, plus output royalties—to foster "fair remuneration" without stifling innovation. OpenAI rejected this overture, but Welp now anticipates "discussions" post-ruling.
OpenAI's response was measured but firm: "We disagree with the ruling and are considering next steps," a spokesperson told Reuters, reaffirming the company's respect for IP while negotiating with stakeholders worldwide. An appeal to the Munich Higher Regional Court is likely, potentially delaying enforcement for months. Yet, the decision has ignited fervent debate on social media, with X (formerly Twitter) users hailing it as a "wake-up call" for AI ethics. One post from AI policy analyst @aifray called it the "first copyright ruling against OpenAI worldwide," predicting ripple effects for models like Google's Gemini and Meta's Llama. Industry watchers, including the 4iP Council, note it clarifies EU law on whether AI training and generation "reproduction rights," potentially reshaping data deals.
For creators, the win is bittersweet. Grönemeyer, whose lyrics fueled the case, has publicly decried AI as a "threat to human artistry" in a 2025 Der Spiegel interview, fearing it devalues emotional labor. GEMA's Holzmüller envisions a future where AI royalties mirror streaming payouts—projected at €1.2 billion for German artists in 2025—ensuring sustainability amid AI's rise. Critics, however, warn of overreach: Tech advocates argue strict licensing could hobble open-source AI, stifling breakthroughs in education and medicine.
As appeals loom, this Munich moment underscores a transatlantic divide: While U.S. courts lean toward "fair use" flexibility, Europe's emphasis on moral rights prioritizes creators. With the EU probing AI data practices under GDPR in parallel probes, OpenAI—and its peers—face mounting pressure to pivot from "scrape-and-train" to "license-and-innovate." For now, GEMA's victory stands as a clarion call: In the AI gold rush, not all data is fair game.
