Zhou Shen's New Song 'Chronicle of the Moon' Released: Clearly Marked as Strictly Prohibited for AI Training and Voice Imitation

marsbit發佈於 2026-04-02更新於 2026-04-02

文章摘要

On April 1, Chinese singer Zhou Shen released the theme song "Yue Zhi Ji" for the fantasy drama "Yue Lin Qi Ji," with a notable copyright declaration explicitly prohibiting the use of the work for artificial intelligence (AI) training. This marks the first case in China where such a restriction was imposed at the time of a song’s release, setting a precedent for protecting vocal rights and creative sovereignty in the digital age. The statement specifies that unauthorized use—including covers, re-recording, remixing, or distribution—is strictly forbidden. It explicitly bans the use of the work in AI training, imitation, learning, or generation activities, addressing growing concerns over AI voice replication and algorithm-driven music plagiarism. Zhou Shen has previously emphasized that while AI can achieve technical precision, it cannot replicate the emotional depth and artistic essence of human performance. As AI music commercializes by 2026, this proactive measure establishes a legal and technical framework to safeguard original vocal works and clarifies authorization standards for AI training data. Industry experts believe such declarations will simplify copyright enforcement by establishing intent at the source. Leading artists like Zhou are helping define ethical and legal boundaries for human-AI collaboration, reaffirming that irreplaceable human emotion remains central to artistic creation.

On April 1, renowned Mandopop singer Zhou Shen released the theme song 'Chronicle of the Moon' for the ancient costume fantasy drama 'Moon Scale Chronicles,' with a clear copyright notice in the song's introduction and lyric sheet stating 'Prohibited for Artificial Intelligence (AI) Training.' This move marks the first case in China where copyright red lines were drawn against AI technology at the time of the work's release, signaling that musicians have entered a new phase of 'digital defense' in protecting their voice rights and creative sovereignty.

The notice specifies in detail: Without the written permission of the copyright owner, any use or distribution (including covers, re-recordings, remixes, etc.) is strictly prohibited; without authorization, the work is strictly forbidden for use in AI training, imitation, learning, generation, and other activities. This directly addresses the increasingly rampant phenomena of 'AI voice cloning' and 'algorithmic song washing' in the industry. Zhou Shen has previously expressed a rational perspective on AI technology in public, emphasizing that while AI can achieve extreme precision through algorithms, it cannot replicate the 'vivid emotions' and artistic soul honed through repeated refinement in human singing.

In 2026, as AI music enters commercial exploration, this 'hardcore' notice not only builds a technical firewall for original voices but also provides a standardized paradigm for addressing the authorization legitimacy of AI training data. Industry experts believe that this approach, which locks in intentional infringement from the source of the work, will significantly lower the threshold for evidence in subsequent copyright enforcement. With leading artists taking the initiative, the music industry is accelerating the construction of legal consensus on the boundaries of human-machine collaboration, reaffirming that irreplaceable human emotion remains the core moat of artistic creation amid the wave of technological advancement.

相關問答

QWhat is the title of Zhou Shen's new song and what specific copyright declaration does it carry?

AThe title of Zhou Shen's new song is '月之纪' (Yue Zhi Ji), and it carries a copyright declaration that explicitly prohibits its use for AI training and voice imitation.

QWhy is Zhou Shen's copyright statement for '月之纪' considered a landmark case in China?

AIt is considered a landmark case because it is the first instance in China where a copyright statement specifically prohibiting use for AI training was issued at the time of a work's release, setting a precedent for digital copyright protection.

QAccording to the article, what specific uses of the song are strictly prohibited without written permission?

AWithout written permission, it is strictly prohibited to use the song in any way, including cover versions, re-recording, remixing, or for AI training, imitation, learning, and generation activities.

QWhat does Zhou Shen believe is the key difference between AI-generated music and human performance?

AZhou Shen believes that while AI can achieve extreme precision through algorithms, it cannot replicate the 'vivid emotion' and artistic soul that comes from the meticulous polishing of a human performance.

QWhat broader industry impact does the article suggest Zhou Shen's declaration could have?

AThe article suggests that this declaration sets a standardized model for the legitimacy of AI training data authorization and helps build a legal consensus on the boundaries of human-AI collaboration, reinforcing that irreplaceable human emotion is the core of artistic creation.

你可能也喜歡

Rubin Ultra大减配,英伟达也扛不住内存涨价了?

知名投研机构SemiAnalysis近日报告指出,英伟达已向主要客户预览了其顶级AI芯片Rubin Ultra的最新规格,但相比此前预期出现显著降配。核心变化包括:算力峰值保持与普通版Rubin相同的35 PFLOPs;显存容量降至8层堆叠的192GB,甚至低于普通版的288GB;显存带宽仅微升1TB/s;芯片功耗反而略有提高。Rubin Ultra的主要升级方向转向“扩展互联规模”,支持通过NVLink将最多576张GPU互联成统一计算域,远高于普通版的72张。 报告分析认为,此次调整主要源于HBM(高带宽内存)价格持续快速上涨。以HBM3为例,其价格已从2025年二季度的低点180-220美元,飙升至目前约700-850美元。HBM成本飙升使得Rubin Ultra单机架的物料成本一度从约660万美元升至800万美元。英伟达因此重新评估设计,通过减少昂贵的HBM配置(成本占比从近40%降至28%),将资源更多投入互联能力等方向,优化整体成本结构。 此消息引发市场对HBM需求见顶的担忧。受冲击影响,韩国存储股开盘大跌,SK海力士、三星股价均下挫约8%。市场解读认为,若英伟达此举成为趋势,意味着AI芯片厂商可能开始通过优化设计来降低对高容量HBM的依赖,这或将限制存储厂商未来的提价空间,标志着AI基础设施的“堆料涨价”时代可能接近尾声。

Odaily星球日报58 分鐘前

Rubin Ultra大减配,英伟达也扛不住内存涨价了?

Odaily星球日报58 分鐘前

生成模型也能端到端训练了?核心竟是一个for循环

长期以来,深度学习在图像分类、检测等任务上凭借端到端训练取得了巨大成功,但生成模型领域却始终是个例外。当前主流的自回归、扩散模型等训练时只学习预测“一小步”,推理时则需要反复展开数百上千步,这种训练与推理的不一致导致了“暴露偏差”等问题,使得生成模型一直未能实现真正的端到端训练。 最近,一篇来自UIUC与哈佛大学的论文提出了“探索式建模(Explorative Modeling,XM)”新范式,其核心是一个简单到近乎朴素的for循环:在每个训练步骤中,模型生成K个候选样本,然后仅选取其中最接近真实数据的一个样本来计算损失和回传梯度。这种做法巧妙地绕开了传统生成模型中因使用重构损失(如平方误差)而导致的“模态模糊”问题——即模型倾向于输出多个可能答案的平均值,而这个平均值往往不属于任何真实的数据模式。 XM通过这种方式,在不拆分生成过程的前提下,直接提升了模型的“生成表达力”,即模型捕捉数据分布中多个独立模式的能力。论文将“探索”验证为继模型参数量、数据量之后的第三根可扩展轴。实验表明,随着模型规模、数据量和计算量的增大,探索带来的性能收益越发显著,在图像、视频、语言等多种任务上均能带来效率与质量的提升。 更重要的是,当探索程度足够时,XM可以实现真正的端到端生成。在机器人控制等任务中,仅需一次网络前向传播的XM策略,其性能可媲美甚至超过需要上百次前向的扩散模型策略,实现了推理效率的飞跃。 尽管“最佳K样本选择”的思想并非全新,但该研究的贡献在于清晰地揭示了这一简单机制如何直接解决生成模型的核心瓶颈,并为实现高效、端到端的生成模型开辟了新的可能性。随着模型规模的持续扩大,探索这一新维度可能将发挥越来越关键的作用。

marsbit1 小時前

生成模型也能端到端训练了?核心竟是一个for循环

marsbit1 小時前

交易

現貨
活动图片