Google’s DeepMind and YouTube collaborated on an ambitious AI music generation tool called Orca that was ultimately shelved in 2023 due to copyright concerns, according to four anonymous sources familiar with the project. The tool represented a significant leap in AI-generated music capabilities, allowing users to create authentic-sounding songs with simple prompts including artist names, lyrics, and musical genres.
Orca’s capabilities were described as “mind-blowing” by insiders, with the ability to generate songs in any artist’s voice—such as creating a hip-hop track using Taylor Swift’s voice. The project was developed as part of Google’s efforts to catch up with OpenAI in the rapidly evolving artificial intelligence race. The tool’s sophistication was so advanced that some employees reportedly suggested changing its code name from “Orca” (a killer whale) due to concerns about its potential to disrupt the entire music industry.
Google attempted to negotiate with major music labels to release Orca publicly, proposing revenue-sharing agreements for the music and artists the AI had trained on. However, the labels rejected these overtures, viewing the technology as a “huge legal risk.” This forced Google to halt the project despite its promising technical achievements throughout 2023.
The project reveals how tech companies have bent their own rules in the AI competition. Google had previously maintained a policy against using copyrighted videos for AI training, but reversed this stance after OpenAI began scraping YouTube for its own models. While Google’s terms allow data scraping from YouTube to improve its services, it remains unclear whether building an AI music generator falls within this policy framework.
DeepMind eventually released a scaled-down version called Lyria in November 2023, which was far more limited in scope. Unlike Orca, Lyria only allowed users to generate music using voices and styles of artists who had explicitly partnered with Google, such as John Legend. Google researchers had also developed a similar model called MusicLM, trained on unlabeled music datasets. Some employees who worked on both Orca and Lyria later departed to found Udio, a startup creating AI music-creation applications. Google did not respond to requests for comment on the matter.
Key Quotes
mind-blowing
An insider described Orca’s capabilities this way, specifically referencing its ability to generate songs like a hip-hop track in Taylor Swift’s voice. This characterization underscores how advanced the technology had become before being shelved.
huge legal risk
A source familiar with the project used this phrase to explain why music labels rejected Google’s revenue-sharing proposal for Orca. This quote captures the fundamental copyright concerns that ultimately killed the project despite its technical promise.
some employees suggested that giving it a code name after a killer whale wasn’t a good idea if DeepMind had the potential to destroy the music industry
This statement from someone involved in the project reveals internal concerns about Orca’s disruptive potential. It demonstrates that even Google employees recognized the existential threat the technology posed to traditional music creation and distribution.
Our Take
Orca represents a watershed moment in AI development where technical capability has outpaced legal and ethical frameworks. Google’s willingness to reverse its own copyright policies reveals how competitive pressure from OpenAI is forcing even the most established tech companies to take risks they previously avoided. The music industry’s firm rejection suggests that creative sectors have learned from past disruptions and are drawing lines early. What’s particularly telling is that the technology was compelling enough that employees left to commercialize similar tools through startups like Udio—indicating that while Google hit a wall, the genie is already out of the bottle. This pattern will likely repeat across other creative industries as generative AI continues advancing, forcing urgent conversations about fair compensation, consent, and the future of human creativity in an AI-augmented world.
Why This Matters
This story illuminates the fundamental tension between AI innovation and intellectual property rights that is reshaping the creative industries. The shelving of Orca demonstrates that even tech giants like Google cannot simply bulldoze through copyright concerns, despite their competitive pressures to match rivals like OpenAI. The music industry’s resistance to Orca signals that creative sectors are drawing firm boundaries around AI training on copyrighted material, potentially setting precedents for other industries.
The revelation that Google reversed its own policies on using copyrighted content shows how the AI arms race is pushing companies to compromise their principles. This has broader implications for how AI models are trained and what data sources are considered acceptable. The fact that some Orca developers left to form Udio suggests the technology’s commercial potential remains strong, even if major platforms face legal obstacles. As AI-generated content becomes more sophisticated, regulators, courts, and industry stakeholders will need to establish clearer frameworks balancing innovation with creator rights—a challenge that will define the next phase of AI development.
Recommended Reading
For those interested in learning more about artificial intelligence, machine learning, and effective AI communication, here are some excellent resources:
Recommended Reading
Related Stories
- Google’s Gemini: A Potential Game-Changer in the AI Race
- Video game voice actors vote to allow use of AI voices
- Jenna Ortega Speaks Out Against Explicit AI-Generated Images of Her
- Photobucket is licensing your photos and images to train AI without your consent, and there’s no easy way to opt out
- OpenAI’s Valuation Soars as AI Race Heats Up
Source: https://www.businessinsider.com/google-deepmind-orca-ai-music-tool-copyright-isues-2024-11