Mitek-owned biometric authentication developer ID R&D is introducing a new feature that can verify the identity of the speaker while they are speaking to voice-enabled chatbots powered by generative artificial intelligence language model ChatGPT.
The IDVoice for ChatGPT feature was demonstrated this week at the natural language and generative AI conference Voice & AI in Washington, D.C.
“Verbal communication with chatbots will soon become commonplace, and securing chatbot sessions from unauthorized access – without burdening users with added friction – will make chatbots that much more useful for a variety of applications that warrant added security,” says Alexey Khitrov, CEO and co-founder of ID R&D.
While ChatGPT is most well-known for its capabilities in generating written content, it can also create voice-enabled content. Companies are currently experimenting with introducing the chatbot, developed by OpenAI, into virtual assistants, voice control and speech-to-text applications, generating content such as audiobooks and more.
The company’s IDVoice performs speaker verification and liveness detection in the background while the user is talking. When a previously enrolled user speaks to ChatGPT, their speech is biometrically authenticated and converted to text. The text is then entered as chat text in the ChatGPT mobile app. If an unauthorized person attempts to communicate with the bot, IDVoice will reject them, the company explains in a release.
The New York-based ID R&D creates voice biometrics software that can be integrated into mobile, web, messaging, and telephone channels, as well as smart speakers, set-top boxes, and other IoT devices.
This August, the company won top spots at the VoxCeleb 2023 Speaker Recognition Challenge in the detection cost function (DCF) and equal error rate (EER) category.