Tencent Music Removed Over 250,000 Songs in 2025 Due to "Emerging AI Risks"

Tencent Music took down over 250,000 songs and reviewed 600,000+ ‘high-risk’ copyright cases in 2025 amid ’emerging AI risks’

May 7, 2026 By Murray Stassen

Tencent Music Entertainment , China’s largest music streaming service provider, says it took down more than 250,000 policy-violating songs and reviewed over 600,000 cases involving “high-risk copyright content” across its platforms last year.

The figures come as TME said it bolstered compliance and risk management across key areas in 2025, including copyright licensing, emerging AI risks, and its overseas business expansion.

The figures were disclosed in TME’s 2025 Environmental, Social and Governance (ESG) Report, published in April.

The 250,000 -plus songs removed from platforms including QQ Music were identified through a combination of AI-powered detection tools and manual inspection as posing “reputational risks” or violating platform content policies.

Separately, TME said it took down over 27,000 songs specifically involved in what it categorizes as “ song theft ,” “ song laundering ,” and “ trend hijacking ” — three distinct forms of so-called “gray-market” manipulation that the company says are “becoming increasingly covert and complex.”

TME defines “ song theft ” as the tampering with rights ownership information; “song laundering” as the plagiarism or alteration of existing music works; and “trend hijacking” as speculative or misleading content creation designed to game the platform’s distribution systems.

The company said these practices “undermine the legitimate rights of original creators, negatively affect their income and creative motivation, and mislead users through speculative practices, ultimately disrupting industry order.”

To combat these issues, TME said it has deployed multimodal recognition technologies including audio fingerprinting , voiceprint recognition , melody comparison , and text similarity analysis for continuous monitoring of its catalog.

The company also said it upgraded its internal rulebook for policing inappropriate content, releasing version 2.0 of its Code of Conduct for Inappropriate Content Management in 2025. That framework introduces a tiered control mechanism with scoring standards and graduated response measures.

Notably, TME said it has also introduced a tiered management system for labels, establishing a “health rating” that continuously assesses their compliance performance.

Labels that fail to meet standards face actions “ranging from warnings and mandated rectifications within a specified timeframe to suspension or termination of collaboration,” the report stated.

Beyond its enforcement against traditional forms of content manipulation, TME’s ESG report also outlined the company’s emerging approach to AI-generated content on its platforms.

TME said it now identifies and tags content containing AIGC (AI-generated content), using audio models, and implements actions based on its governance rules. The company said it has established “a regular content security assessment and detection mechanism” for AI-generated content.

On the question of AI model training, a topic of intense debate in the global music industry, TME stated that it adheres to the principle of “legally compliant licensing,” and that it only incorporates content into AI model training and creative scenarios “after obtaining licensing from copyright holders.”

The company added that it “clarif[ies] responsibilities regarding content usage and authorization through AI-specific clauses in product user service agreements” and collaborates with music labels “to deploy technology that identifies and addresses unauthorized AI-generated content.”

TME’s disclosure adds another major platform to the growing list of streaming services grappling with AI-generated content.

Deezer , which says it became the first streaming platform to explicitly tag AI-generated music when it launched its detection system in June 2025, said in April 2026 that it now receives nearly 75,000 fully AI-generated tracks per day — roughly 44% of all daily uploads — and tagged more than 13.4 million AI tracks across 2025.

Apple Music introduced AI Transparency Tags in March 2026, a set of metadata labels that labels and distributors can apply now and will be required to use when delivering new content in future, though the system relies on self-reporting.

Spotify launched an AI Credits beta feature in April 2026, built on the DDEX metadata standard, allowing artists to disclose AI use in specific creative contributions such as vocals, instrumentation, or post-production. Spotify has also removed over 75 million tracks it classified as “spammy” and, at the end of April, introduced a “Verified by Spotify” badge to help distinguish human artists from AI-generated personas.

TME’s copyright enforcement efforts sit within a broader intellectual property portfolio that, as of December 31, 2025, included 5,530 patent applications ( 3,513 granted), 6,544 trademark applications (4,570 registered), and 741 registered software copyrights.

The company’s patent filings in 2025 focused on generative AI technology, intelligent recommendation, audio recognition, and audio quality and sound effects.

Meanwhile, TME itself faces a significant volume of copyright claims. According to the company’s 2025 annual report, published via the US SEC last month, there were 160 lawsuits pending in connection with alleged copyright infringement on its platforms as of December 31, 2025, with an aggregate amount of damages sought of approximately RMB 187.1 million ( USD $26.8 million ).

TME said it does not believe any such proceedings are likely to have a material adverse effect on the company.

The company also disclosed that it entered into a strategic copyright cooperation agreement with the Music Copyright Society of China (MCSC) during the year, focused on protecting the copyright in musical lyrics and compositions and optimizing licensing mechanisms.

TME said the partnership combines MCSC’s capabilities in copyright management and enforcement with its own platform’s technological capabilities, with the aim of promoting “a more standardized, transparent, and efficient licensing process for lyrics and compositions”.

The company added that the collaboration “provides creators with clearer licensing pathways and a more reliable revenue guarantee mechanism.”

TME also said that, as a member of the Copyright Society of China , it co-initiated the establishment of an Artificial Intelligence Copyright Working Committee , where it serves as deputy director. The committee engages in research and discussions on emerging issues related to the copyright of AI-generated works.

On the international stage, TME said its copyright protection practices were showcased as exemplary cases at the World Intellectual Property Organization (WIPO) General Assembly meetings, where the company contributed insights from its governance of the Chinese digital music market.

TME also participated in the National Conference on Copyright Protection and Development in Digital Environment, discussing trends toward stronger copyright compliance in the digital music sector, and took part in the China Digital Music Industry Conference, engaging in dialogue on topics including technology empowerment, global content distribution, and developments in AI.

In terms of standard-setting, TME said it took the lead in drafting the Cybersecurity Standard Practice Guidelines : Labeling Method for Content Generated by Artificial Intelligence – File Metadata Implicit Label for Audio Files, contributing to the establishment of labeling norms for AI-generated audio content.

In collaboration with the China Audio Industry Association , the company also released two audio-related group standards covering device playback capability for high-quality source audio and spatial audio interaction between platforms and wireless earphones. Music Business Worldwide

_Originally reported by [Music Business Worldwide](https://www.musicbusinessworldwide.com/tencent-music-took-down-over-250000-songs-and-reviewed-600000-high-risk-copyright-cases-in-2025-amid-emerging-ai-risks/)._