OpenAI has unveiled a powerful new tool called Voice Engine that can generate highly realistic synthetic speech from just a 15-second audio sample. However, the AI research company is taking a cautious approach to its broader release due to concerns over potential misuse, especially during election cycles.

Acknowledging the "serious risks" of voice impersonation technology, OpenAI stated it is engaging with government, media, entertainment, education, and civil society partners to gather feedback and implement proper safeguards before unleashing Voice Engine more widely.

The careful rollout comes after recent incidents involving AI-generated voice deepfakes being used for disinformation, such as a robocall impersonating President Biden's voice urging voters not to participate in a primary election.

Experts warn that cheap, easy-to-use voice cloning tools could enable a deluge of audio-based misinformation during the 2024 US presidential race and other key global elections this year.

To mitigate risks, OpenAI has implemented watermarking to trace generated audio, monitoring usage, and requirements for partners - including obtaining explicit consent before cloning someone's voice and disclosing when voices are AI-generated.

While exploring Voice Engine's potential benefits, OpenAI's measured rollout prioritizes implementing sufficient guardrails against nefarious exploitation of the powerful voice synthesis capability.