Is that song AI-generated? UChicago scientists create tool to check

Is that song AI-generated? UChicago scientists create tool to check

‘Quicksilver’ browser extension intended to champion transparency and ethics

In 2024, a remarkable shift hit the music world: AI-generated songs exploded onto streaming platforms, making up nearly half of all newly uploaded tracks. But as the lines blur between human artistry and algorithmic output, listeners face a new question: How much of the music we truly enjoy is made by people, and how much is the work of machines?

A new cutting-edge browser extension from a team of scientists at the University of Chicago, known as Quicksilver, allows listeners to scan songs for traces of AI.

The project is intended to champion transparency, ethical technology, and support from human creativity in a rapidly evolving digital soundscape, its authors said.

Quicksilver was developed by the UChicago SAND Lab alongside the nonprofit Ethical Technology and Computing for Humanity (ETCH), launched this year by UChicago Neubauer Professors of Computer Science Ben Zhao and Heather Zheng with a mission to ensure technology serves the broader interests of society and creative communities.

It follows Glaze and Nightshade, recent work by Zhao and Zheng that disrupt nonconsensual training on artistic work; these two programs have now been downloaded over 13 million times by creatives in more than 160 countries.

“Given that there is so much AI music out there, and that normal users can’t tell the difference, giving users a tool to identify these AI songs was a very natural solution,” said Stanley Wu, lead developer and UChicago graduate student. “At this scale, I think there needs to be more protections in place so that this wave of “spammy” AI music does not negatively impact human artists.”

A tool built for transparency and privacy

As members of the SAND Lab team measured the prevalence of AI music on platforms like Spotify and Apple Music, they discovered that most platforms don’t disclose whether tracks are AI-generated or not.

Their research revealed that close to 50% of weekly releases are now AI, most receiving little engagement—a wave of so-called “AI slop.”

Wu and his colleagues, including Zhao, Zheng, and undergraduates Naryna Azizpour and Viresh Mittal, found the problem only deepened the more they investigated: Even professional musicians performed only slightly better than random chance at distinguishing AI music from human compositions.

With AI music so prolific, often created with no human input and little engagement, questions about artist integrity and impact on real musicians have become urgent, Wu said.

Unlike many detection services, Quicksilver operates locally on your device. Users play music from any streaming platform and Quicksilver listens in, analyzing the audio live.

There are subtle “artifacts” in AI-generated audio that ears simply can’t detect, according to the team. With a tap of the “Analyze” button, the extension scans for these artifacts, particularly those produced via Suno and Udio, two widely used AI music platforms.

No audio is uploaded to external servers, and detection is lightweight and fast.

Musicians’ responses to Quicksilver have been overwhelmingly positive, and the team received supportive feedback from the French streaming platform Deezer, whose research also inspired Quicksilver’s detector.

The team is actively working to keep pace with newer music generation models and evolving deepfake technologies. Their ongoing priority is to ensure that Quicksilver remains accurate and responsive, minimizing false positives as it adapts to new challenges.

A nonprofit for ethical development of new technologies

Quicksilver’s roots reach back to ETCH, a nonprofit spun out from years of SAND Lab research. ETCH includes board members and advisors from across UChicago’s Computer Science, Law, Medical, and Booth School of Business, and is dedicated to investigating and guiding the ethical development of new technologies.

The organization’s philosophy is clear: technology must serve humanity. “Human creative work represents the very best of human expression and communication,” ETCH’s website notes. “Creative forms of human expression should be celebrated, preserved, and protected by technology.”

ETCH’s funds and supports ethical research, partners with technologies and creators, translates complex findings into actionable guidance, and builds resources to help creative communities navigate technological change.

“With trillions of dollars committed to developing, deploying, and monetizing AI systems in nearly all aspects of our lives, it is more important than ever to elevate the human voice, to highlight the value of human creativity and ingenuity even as we explore ethical and principled approaches towards AI,” said Zhao. “The goal of ETCH is to ensure that AI efforts are transparent, accountable, and equitable while elevating human creativity and prioritizing social good over profit.”

Looking forward, the researchers are hopeful that industry stakeholders will embrace tools like Quicksilver to promote greater transparency around AI-generated music and streaming content.

“The continued development of AI is poised to change society in ways few of us can predict,” said Zhao. “We believe that technology should always serve humanity, and it is up to all of us to ensure that AI becomes a benefit for the many, not the few. At ETCH, we are working towards this goal in multiple creative domains while engaging with stakeholders across the ecosystem.”

—Adapted from an article first published by the Department of Computer Science.

Big Brains podcast: Fighting back against AI piracy

What makes music go viral—From AI to Taylor Swift