OpenAI Recommends Phaseout of Voice-Based Authentication While Previewing Text-to-Speech Model

OpenAI suggested Friday (March 29) that voice-based authentication for accessing bank accounts and other sensitive information should be phased out. 

The company said this while announcing in a blog post that it has developed and previewed but not widely released an artificial intelligence (AI) model that can generate speech closely resembling that of an individual. To do so, it needs only text input and a 15-second audio sample of that individual. 

OpenAI also suggested in the post that society should explore policies to protect the use of individuals’ voices in AI, educate the public about the possibility of deceptive AI content, and accelerate the development of ways to determine whether a speaker is a real person or AI. 

“It’s important that people around the world understand where this technology is headed, whether we ultimately deploy it widely ourselves or not,” OpenAI said in the post. 

The company plans to “start a dialogue” on the responsible deployment of synthetic speech before releasing this Voice Engine model, according to the post.

“Based on these conversations and the results of these small scale tests, we will make a more informed decision about whether and how to deploy this technology at scale,” the post said.

OpenAI has shared the Voice Engine text-to-speech model with about 10 developers but scaled back its previously announced plan to release it to as many as 100 developers, Bloomberg reported Friday, citing a company spokesperson. 

The company dialed back its plans after getting feedback from policymakers, industry experts, educators, creatives and other stakeholders, according to the report.

In its Friday blog post, OpenAI said: “We recognize that generating speech that resembles people’s voices has serious risks, which are especially top of mind in an election year.”

The company’s partners who are testing Voice Engine have agreed not to impersonate people or organizations without their consent, and to disclose to their audiences that the voices are AI-generated, according to the post.

OpenAI has implemented watermarking to trace the origin of any audio generated by the tool and it proactively monitoring how the tool is being used, per the post.

“We believe that any broad deployment of synthetic voice technology should be accompanied by voice authentication experiences that verify that the original speaker is knowingly adding their voice to the service and a no-go voice list that detects and prevents the creation of voices that are too similar to prominent figures,” the post said. 

The growing interest in and adoption of voice technology has generated concerns about potential misuse for fraudulent activities, PYMNTS reported Friday. 

In February, the Federal Communications Commission (FCC) made it illegal to use AI-generated voices in robocalls, saying its ruling will give state attorneys general another tool to use against voice cloning scams.

Federal regulators have already recorded a jump in the use of all forms of AI in scams. On March 17, PYMNTS reported that the Federal Trade Commission (FTC) has seen a sharp rise in complaints about ads that either used AI or claimed to use it to pull people into scams.