Bold claim: Google’s NotebookLM is under fire as a well-known radio host accuses the tech giant of stealing his voice. But here’s the real story and the nuances you need to know.
Former NPR anchor David Greene has filed a lawsuit against Google, alleging that NotebookLM’s Audio Overviews feature uses a male AI voice that is an unauthorized replica of his own vocal style. Greene contends that the AI mimics his distinctive cadence, intonation, and even habitual fillers like “uhhs” and “likes” that he has spent years refining to minimize, but could not eliminate. He argues that this replication was achieved without proper payment or permission and that users can prompt the system to generate speech Greene would never say.
Greene’s broadcasting career, notably on Morning Edition from 2012 to 2020, reached approximately 13 million listeners, making it one of the most prominent news programs in the United States. He has since moved on to host the political podcast Left, Right & Center. He asserts that Google trained NotebookLM on his voice without consent, a claim Google rejects as baseless. The tech giant maintains that the Voice in Audio Overviews comes from a professional voice actor hired for the project, not from Greene’s recordings.
Google emphasizes that the voice used in NotebookLM was contracted to a paid actor, not derived from Greene’s voice. The company’s stance aligns with prior industry practices, where synthetic voices are produced by hired talent rather than duplicating a public figure’s likeness without authorization.
This case isn’t isolated in the AI space. In 2024, Scarlett Johansson publicly challenged OpenAI over a ChatGPT-4o voice named “Sky” that resembled her own. OpenAI reported that the voice was not a deliberate imitation, and the voice-over talent had been engaged months before Johansson’s involvement, highlighting the complexity of voice likeness in AI training.
Key takeaway: As AI voice capabilities grow, questions about consent, attribution, and fair use become increasingly prominent. This dispute could influence how companies train and deploy voice models, especially for consumer-facing features like Audio Overviews and Video Overviews.
What do you think? Should companies be allowed to create synthetic voices that resemble real people if they hire professional voice talent to do the work, or should explicit permission and compensation be mandatory for any likeness used in AI products? Share your thoughts in the comments.
