On Monday, ElevenLabs – a venture led by ex-Google and Palantir employees – reported a sharp rise in the misuse of their newly launched beta. After further investigation, Motherboard discovered that some 4chan members had employed the product to generate voices resembling popular personalities such as Joe Rogan, Ben Sharpio, and Emma Watson – to spread hateful rhetoric laced with racism and other forms of prejudice. In light of this, ElevenLabs has committed to exploring stricter safety measures regarding their technology.
The audio clips circulating on 4chan are mainly centered around famous individuals. With excellent voice replication quality and its accessibility, these deepfake audio clips present an imminent issue. Deepfake videos were first utilized for creating non-consensual porn featuring certain subjects. Now, deepfake audio is following an eerily similar course. The content ranges from violent to transphobic, homophobic, and racist messages. One 4chan post also included a link to ElevenLabs’ beta, potentially indicating that the software was used to fabricate these voices.
In light of this, the tech firm is actively exploring further safeguards for its technology. The implications of realistic deepfake audio are just beginning to unfold.