Experts worried

Program from OpenAI can now clone voices

Due to the potential for the misuse of synthetic voices, the company has opted for a "cautious" approach to wider release, it says.(Bild: ArtemisDiana - stock.adobe.com)

The ChatGPT developer OpenAI presented a program on Friday that can clone voices. Experts now fear that the new artificial intelligence capabilities are being misused - especially in light of this year's EU and US presidential elections.

The model called "Voice Engine" can duplicate the voice of a human based on a 15-second audio original, according to a blog post by OpenAI, which presents the results of a test with the program.

OpenAI stated that it would take these problems into account. The San Francisco-based company explained that it is aware that the generation of voices that resemble those of humans harbors serious risks that require particular attention in an election year. The company is working with partners from the fields of politics, media, entertainment, education and civil society, among others, and their feedback will be taken into account during development.

(Bild: ymgerman - stock.adobe.com)

Rules are being developed
Due to the potential for the misuse of synthetic voices, the company has opted for a "cautious" approach for a broader release. Partners who tested "Voice Engine" had agreed to rules. Among other things, these require the express consent of every person whose voice is duplicated with the program. It must also be made clear to listeners that the voices they hear have been generated by AI.

Artificial intelligence is suitable for the masses
OpenAI initially caused a stir in November 2022 with the launch of ChatGPT. The program is able to use artificial intelligence (AI) to generate essays, poems or conversations from very short prompts, for example. ChatGPT suddenly made a large audience aware of the possibilities of AI. At the same time, however, fears grew about the potential dangers of the technology.

The company, which also developed the image generator Dall-e, also presented an AI application for producing realistic videos called Sora in February.