CHANG TSI
Insights
In April 2026, Taylor Swift, through TAS Rights Management, filed with the USPTO two trademark applications for exclusive spoken-phrase sound marks and one trademark application for a tour-look image mark, in an effort to use trademark registration as a preemptive measure to intercept AI-generated deepfake voices and unauthorized endorsements. As generative AI continues to disrupt the global entertainment and commercial landscapes, corporate legal departments and practitioners are facing an urgent question: Can the Western sound trademark blueprint protect commercial and individual voice assets within China’s jurisdiction, and what fallback strategies must be deployed when trademarks hit their limits?
Under Article 8 of China Trademark Law, sound is explicitly recognized as a registrable trademark, permitting legal protection for exclusive auditory identifiers that exhibit distinctiveness and structural stability. While historic registrations, such as the Tencent QQ chime and Intel’s corporate sound effect, successfully empower brands to enjoin third parties from copying identical audio for competing commercial uses, sound trademarks offer highly localized defense lines. They cover only fixed, registered audio snippets. Consequently, trademarks fail to safeguard an individual's unformatted native vocal timbre, pitch, or fluid vocal expressions, leaving a protection gap that fails to contain multi-dimensional AI voice cloning threats.
When trademarks fall short, the China Copyright Law serves as a vital secondary safety net. Under Articles 10, 39, and 44, copyright owners of compositions hold exclusive reproduction and information network dissemination rights, while performers and audio producers maintain statutory exclusivity over their live executions and master sound recordings. This statutory framework effectively penalizes platforms or developers that scrape original master files without authorization to train proprietary AI models or commercialize AI-synthesized covers. Because copyright focuses heavily on fixed artistic expressions, it provides a robust legal avenue to combat the unauthorized exploitation of original acoustic materials and secondary adaptations.
For abstract AI infringements—such as mimicking a specific vocal tone without replicating a copyrighted work or a registered trademark—rights holders can build an aggressive defense using the PRC Civil Code and Anti-Unfair Competition Law. Article 1023, Paragraph 2 of China's Civil Code expressly provides that the protection of a natural person's voice shall apply, by reference, the rules governing the protection of the right of portrait, thereby providing a direct legal basis for the judicial protection of voice-related interests. This is a doctrine solidified by the Beijing Internet Court in the landmark Yin Case (2024). The court ruled that an AI-synthesized voice infringes upon personality rights if its unique timbre and intonation are publicly recognizable as mapping to a specific individual. Crucially, the precedent clarified that purchasing a master recording’s copyright does not authorize AI training, which constitutes a distinct form of exploitation requiring explicit, independent and specialized consent. Furthermore, if synthetic voices induce commercial confusion, they trigger Article 6 of the Anti-Unfair Competition Law, which penalizes the unauthorized use of influential identifiers to mislead consumers.
To navigate this landscape, a tiered, profile-specific compliance framework should be adopted:
• For Enterprise Brands: The primary objective is shielding standardized corporate auditory identifiers rather than human vocal traits. Companies should secure sound trademarks for long-term, high-value promotional slogans and sound effects, maintain rigorous copyright repositories for all corporate audio, and embed explicit restrictive covenants in vendor contracts prohibiting the extraction of brand audio for AI model training.
• For Public Figures and Voice Talent: Protecting an elite vocal profile requires a multi-layered legal matrix spanning fixed clips, original performances, and natural vocal timbre. While trademarks and copyrights deter direct asset theft, unauthorized AI endorsements must be aggressively countered via the Civil Code's voice personality rights. Rights holders must proactively secure public recognition evidence through notarization and blockchain timestamps, while layering anti-unfair competition arguments to halt market dilution.
• For AI and Content Platforms: Platforms must implement strict pre-clearance verification alongside responsive post-infringement takedown workflows to mitigate secondary liability. Platforms are obligated to verify written chain-of-title documentation ensuring that users uploading vocal assets possess explicit rights covering AI training, synthesis, and commercial distribution, while preserving end-to-end content generation and transmission logs.
It's worth noting that Taylor Swift is not the first public figure to attempt using a sound trademark to counter AI cloning: American actor Matthew McConaughey previously filed a sound trademark application with the USPTO for his signature catchphrase "Alright, alright, alright!", which was approved in December 2025, making it an early precedent in this area. Taylor Swift's filing can be seen as a continuation and escalation of this international trend at the level of top-tier artists, and also indirectly confirms that, in the AI era, sound trademarks are gradually moving from a "niche, obscure category" toward becoming a standard option in the IP protection strategies of public figures.
While Taylor Swift’s forward-looking trademark initiative offers an inspiring case study, it functions primarily as a localized defense line rather than a comprehensive solution to generative AI risks. To achieve resilient, long-term asset security, market participants must move past single-remedy approaches and deploy a synchronized "portfolio strategy" combining trademarks, copyrights, personality rights, and anti-unfair competition law tailored to China's evolving judicial landscape.
