Two major court decisions in July 2026 have created a fundamental divide in how the world treats AI copyright: Germany's Munich Regional Court ruled that AI music training violates copyright law, while India's Delhi High Court approved the same practice under fair use doctrine. These competing rulings expose a growing fracture in global IP law that will force AI companies to operate under radically different legal frameworks depending on geography.

What Did Germany's Court Actually Rule?

On July 31, 2026, the Munich Regional Court handed down a decisive judgment against Suno AI, the popular music generation platform, in a case brought by GEMA, Germany's music collecting society. The court found Suno liable for copyright infringement across its entire operation, from the initial training of its AI model to the music outputs users generate.

The ruling introduced several novel legal concepts that expand copyright liability beyond what US courts have previously accepted. The court established that when an AI system retains the ability to regenerate protected musical elements, those elements are considered "memorized" within the model's parameters. This memorization counts as permanent fixation and reproduction, not a transient analytical step. The court also rejected Suno's fair use defense outright, distinguishing US precedents by noting that Suno's outputs were substantially similar to original compositions when users provided specific prompts.

Critically, the Munich court held that the AI provider, not the user, bears primary liability. While users trigger the system with simple text prompts, Suno controls the model's design, training data, and resulting memorization. The court also noted that Suno obtained training material through "stream-ripping" from YouTube, circumventing YouTube's technical protections. This unlawful acquisition of training data itself constituted a separate violation.

"Compliance with statutory AI safety or transparency duties under the EU AI Act does not provide a defense against copyright infringement, as regulatory compliance cannot substitute for acquiring a copyright license," the court stated in its reasoning.

Munich Regional Court, GEMA v. Suno AI judgment

How Did India's Court Reach the Opposite Conclusion?

Just one week before Germany's ruling, on July 24, 2026, India's Delhi High Court delivered a landmark interim decision in ANI Media v. OpenAI that explicitly protected AI training from copyright liability. The court refused to grant an injunction against OpenAI for scraping news articles to train ChatGPT, establishing a broad pro-AI framework under Indian copyright law.

Justice Amit Bansal's reasoning differed fundamentally from Germany's approach. The court held that ANI Media failed to demonstrate copyright infringement in ChatGPT's outputs, emphasizing that copyright protects specific creative expression rather than underlying factual content. The court also noted that several articles ANI cited were technically impossible for OpenAI to have used, as they post-dated ChatGPT's training cutoff date.

More significantly, the Delhi court recognized machine learning itself as a form of statutory "research" eligible for protection under India's Fair Dealing doctrine. Because internal training extracts patterns in a closed environment without making raw data public, the court found it satisfied the purpose test for private research, a defense available even to commercial entities. The court also emphasized that the balance of convenience disfavored an injunction, noting that OpenAI had voluntarily blocked its crawlers and that requiring individual training licenses would curtail domestic AI development.

"India is considered a forerunner in the AI field and the law must continually catch up with technology," the court stated, underscoring the need for legal frameworks that support innovation.

Justice Amit Bansal, Delhi High Court

What Does This Split Mean for AI Companies?

The divergence between Germany and India creates an immediate practical problem for AI developers: the same training practice that is illegal in one jurisdiction is legal in another. Suno is expected to appeal the Munich decision, and the case will likely reach Germany's appellate courts and potentially the Court of Justice of the European Union (CJEU), which is hearing a parallel case, Like Company v. Google, that will address copyright and AI training at the EU level.

Meanwhile, the US litigation landscape remains unsettled. The Massachusetts federal court overseeing Universal Music Group and Sony's case against Suno reset its schedule on June 30, 2026, pushing dispositive motions to April 9, 2027. This means a US fair use ruling will not arrive until 2027 at the earliest, leaving American AI companies operating in legal limbo.

The music industry has already begun bifurcating into licensed and unlicensed segments. Universal Music Group settled with Udio in October 2025, and Warner settled with both Udio and Suno in November 2025. However, UMG and Sony remain in active litigation against Suno in Massachusetts, and Sony continues to litigate against Udio in New York.

How Are Regulators Responding to the Copyright Question?

The European Union has moved to establish direct copyright governance for AI through the EU AI Act, which entered a key enforcement phase on August 2, 2026. Under Article 53 of the regulation, General Purpose AI (GPAI) providers must maintain a policy to comply with EU copyright law and respect rightsholder opt-outs against text and data mining (TDM) under the Digital Single Market Directive. Providers must also publicly disclose a detailed summary of their training data, granting rightsholders the transparency needed to identify incorporated material and enforce claims in court.

These obligations primarily apply to GPAI providers and foundational model developers. Downstream AI companies that fine-tune base models may also face compliance requirements regarding their proprietary dataset additions. The regulatory approach differs sharply from the court-driven frameworks emerging in Germany and India, suggesting that the future of AI copyright will be determined by a combination of judicial rulings and legislative mandates.

What Are the Key Takeaways for the AI Industry?

• Geographic Fragmentation: AI companies must now operate under fundamentally different copyright rules depending on jurisdiction. Germany treats AI training as copyright infringement; India treats it as fair use research. The US outcome remains pending.
• Licensing Becomes Strategic: Companies like UMG and Warner chose to license rather than litigate, a decision validated by Germany's ruling. Licensing agreements are now a competitive advantage, not just a legal obligation.
• Technical Acquisition Matters: The Munich court's focus on how Suno obtained training data (stream-ripping YouTube) suggests that the method of data collection, not just the use of data, can create independent copyright liability.
• Regulatory Compliance Is Not a Defense: Complying with the EU AI Act's transparency and opt-out requirements does not shield AI providers from copyright infringement liability. Regulation and copyright law operate on separate tracks.
• Output Similarity Still Counts: Germany's court found that when AI outputs are substantially similar to original works, liability attaches. This differs from India's approach, which focused on whether outputs reproduced specific creative expression.

How Should AI Companies Navigate This Uncertainty?

• Implement Opt-Out Mechanisms: Establish systems to respect rightsholder opt-outs against text and data mining, as required by the EU AI Act. This is now a baseline compliance requirement in Europe and may become standard globally.
• Document Training Data Provenance: Maintain detailed records of where training data originated and how it was acquired. Courts are increasingly scrutinizing the acquisition method, not just the use of data.
• Pursue Licensing Agreements: Negotiate licensing deals with major content holders in key markets. The settlements between music labels and Udio and Suno demonstrate that licensing can resolve litigation and provide competitive advantage.
• Monitor Appellate Decisions: Track the outcomes of pending cases in Germany, the EU, and the US. The CJEU's ruling in Like Company v. Google and the US fair use decision in 2027 will likely reshape the global landscape.
• Differentiate by Market: Consider offering region-specific versions of AI products that comply with local copyright frameworks. This may require different training approaches or output restrictions in different jurisdictions.

What Happens Next in the Courts?

Suno is expected to appeal the Munich decision, joining an ongoing appeal in GEMA v. OpenAI before the Munich Court of Appeals. The CJEU's ruling in Like Company v. Google, which will address copyright and AI training at the European level, is also pending. In the US, the Massachusetts court will not issue a fair use ruling until 2027 at the earliest, meaning the American legal framework will remain uncertain for at least another year.

The Philippines and other emerging markets are also grappling with how to balance AI innovation and IP protection. At a recent forum on intellectual property, government officials and experts emphasized the need for updated policy frameworks that combat piracy while enabling AI development. The consensus was that AI should amplify human creativity, not replace it, and that copyright protection remains essential to sustaining creative industries.

"AI is changing the way we innovate, but it does not change why intellectual property exists. Innovation will always begin with human imagination, ideation. AI may accelerate creativity, but it is the people who provide the vision, judgment and responsibility," stated a former IP office director.

Rowel S. Barba, Former Director General, IPOPHL

The global AI copyright landscape is now defined by competing legal theories, pending appellate decisions, and regulatory mandates that operate independently of court rulings. Companies operating in this environment must prepare for a fragmented world where the same technology is legal in one jurisdiction and illegal in another, at least until higher courts and international bodies establish more uniform standards.