In a bid to address the increasing threat of deepfakes during the global election year, OpenAI is working on a text-to-speech AI model called ‘Voice Engine’ that can generate natural-sounding speech with just a 15-second audio sample.
The AI model has the ability to create emotive and realistic voices, raising concerns about the potential misuse of this technology, especially in the context of elections. OpenAI is collaborating with government, media, entertainment, and other stakeholders to ensure responsible development and usage of ‘Voice Engine’.
Partners testing the technology have agreed to strict usage policies, which prohibit impersonation without consent and require explicit, informed consent from the original speaker. Users must also disclose that the voices are AI-generated, and OpenAI has implemented safety measures like watermarking and monitoring to prevent misuse.
