Artificial intelligence (AI) is changing how music is created, distributed and consumed. But as generative AI continues to evolve, artists face a growing challenge: the unintended and unauthorized usage of their work by AI for training and synthetic content creation.
For Jian Liu, Ph.D., founder and CEO of ArtyShield.ai, the answer begins with understanding AI system vulnerabilities and finding ways to turn those vulnerabilities into protective, trustworthy solutions for artistic works.
“Human creativity deserves to be protected,” Liu said. “At ArtyShield we want to build a more sustainable, more trustworthy ecosystem to help artists in the age of AI.”
Founded in August 2025, ArtyShield is developing a suite of technologies designed to help artists, labels, distributors and rights holders protect and verify creative work. Its products include MusicShield, which protects music against unauthorized AI training and manipulation; VoiceShield, which protects vocal recordings from unauthorized AI cloning; and VeriTune and VeriVoice, which help detect AI-generated music and speech.
Liu, an associate professor at the University of Georgia, developed the core technology behind MusicShield at the University of Tennessee, Knoxville (UTK). The technology is licensed to ArtyShield through the UT Research Foundation (UTRF), who continues to support Liu in bringing MusicShield to market.
UTRF helped Liu navigate the transition from researcher to entrepreneur, providing access to UTRF’s Executives in Residence, pitch deck reviews, investor introductions and other resources to help advance ArtyShield.
“I have no experience in starting a company,” Liu said. “UTRF provided great resources to me so I could understand what I should do if I want to build a company on top of a technology I developed at the University.”
From Studying AI Vulnerabilities to Protecting Artists
When Liu began his faculty career at UTK, he studied adversarial machine learning and the vulnerabilities of AI systems used for speech recognition.
This work gave him a different perspective as generative AI rapidly advanced.
Liu and his recent doctoral student, Syed Irfan Ali Meerza, Ph.D., who now serves as ArtyShield’s chief technology officer and an assistant professor at Virginia Commonwealth University, realized the vulnerabilities they had been studying could also be used to protect creative work from unwanted AI exploitation.
While the law provides protection for copyright, NIL, deepfakes and other similar rights, enforcing those rights after unauthorized AI usage can be difficult, time consuming, expensive, and is only retroactive. Liu and co-creator Meerza saw an opportunity to develop a proactive layer of protection to help prevent AI systems from effectively learning from protected works in the first place.
“From my research I have learned the vulnerabilities of AI models,” Liu said. “I know we can use some tiny noises to protect the data from AI exploitation.”
That concept became the foundation for MusicShield.
MusicShield works by adding imperceptible noises to an audio file that interfere with an AI system’s ability to learn from it. The changes are imperceptible to human listeners but make the audio file difficult for AI models to interpret, caption, use in representation learning and repurpose downstream.
“We introduce carefully designed, imperceptible perturbations into the music so that it sounds the same to human listeners but becomes much less useful for AI models to learn from,” Liu explained.
The approach is designed to allow artists to distribute their music as intended, while adding a proactive layer of protection against unauthorized AI training.
ArtyShield recently took a significant step toward bringing MusicShield into the music industry through a partnership with Symphonic Distribution, a global music distribution and technology company. Symphonic clients can now access ArtyShield’s services through SymphonicMS, giving artists a more direct way to incorporate AI protection and detection into their existing workflows.
Protecting Creativity Before It Is Exploited
The need for proactive protection is rooted in a growing concern throughout the creative industries.
Generative AI models can utilize enormous amounts of training data, including music and other creative works that may be available online. When AI companies train models on creative works, the resulting systems can generate new material that resembles the work they learned from.
For musicians, distributing their music to listeners is fundamental and intentional, but creating and releasing a finished recording requires significant time, energy and money. As AI-generated music becomes more prevalent on streaming platforms, Liu believes the growing volume of synthetic content could make it more difficult for human artists to earn a living.
“If AI-generated work is flooding streaming platforms then that will largely decrease the revenues human musicians could make,” he said.
For Liu, the issue isn’t whether AI should exist but instead building a more robust system of accountability.
“There’s nothing wrong with AI models,” he said. “The problem is when AI companies train their models on copyrighted music without permission.”
The company’s goal is not to stop technological innovation, but to give creators more control over how their work participates in an AI-enabled economy, ultimately allowing artists and copyright holders to negotiate for compensation when their work is used as training data for AI.
“I think everyone loves innovation; we just want to ensure the innovation won’t harm a certain group of people,” Liu said. “At the least, we should have a more fair, more sustainable ecosystem to make human and AI-created work coexist.”
What Comes Next
Beyond proactive protection and AI-content detection, ArtyShield is also developing technologies for content fingerprinting, attribution and copyright-infringement detection. These capabilities are designed to help creators and rights holders identify where their work may have been reused, replicated or transformed, and provide stronger technical evidence for protecting their rights.
Another direction is ArtyShield Certify, a certification workflow designed to help artists establish the provenance of their work and provide verifiable evidence that it was created by a human.
“In the age of generative AI, I believe human-created work may become increasingly valuable, and artists should have a trusted way to communicate that value to platforms, partners and listeners,” Liu said.
ArtyShield is not only aiming to prevent AI from training on original work but also creating infrastructure that can help establish where creative work originated, how it was used, and whether it has been altered or replicated.
As AI systems continue to develop, so will the legal and ethical questions surrounding them. Liu believes technology will have to be part of the solution.
“Many technologies need to be integrated into the current ecosystems, and all stakeholders have a responsibility to help build such a sustainable ecosystem,” he said.
For ArtyShield, that means building technology that can help artists navigate an increasingly complex relationship between human creativity and AI.
“I really hope we can be part of this creative ecosystem,” Liu said.