According to a report from the Information, OpenAI is prioritising audio AI development because the technology will be at the core of a much-anticipated line of physical devices, created by legendary iPhone designer Jony Ive. — Reuters
OpenAI recently reorganised several of its teams in order to focus on improving its audio-generation AI models, according to a new report – and these improvements are crucial to OpenAI pulling off one of its most high-profile projects.
According to a report from the Information, OpenAI is prioritising audio AI development because the technology will be at the core of OpenAI’s much-anticipated line of physical devices, created by legendary iPhone designer Jony Ive.
OpenAI bought Ive’s design startup, io Products, for US$6.5bil (RM26.42bil) in May 2025, with the explicit goal of creating a new generation of AI-powered devices. In the months since, rumours have circulated that the devices would eschew screens in favour of an audio-based operating system, in the form of a pair of smart glasses or an Amazon Echo-like speaker.
OpenAI CEO Sam Altman further added to those rumours during a December 2025 interview with journalist Alex Kantrowitz, in which he said that using a screen would limit OpenAI’s device to “the same way we’ve had graphical user interfaces working for many decades.”
During a presentation in summer 2025, according to the Information, researchers working on the devices told OpenAI staff that the company’s initial device will process information about its surroundings through audio and video, and “act like a companion that works alongside its user, proactively giving suggestions to help the user achieve their goals, rather than as a simple conduit to apps and other software.”
In practice, the device sounds remarkably similar to the one featured in the 2015 film Her. In that film, humans wear devices that allow their AI companions to see and hear the world, while providing commentary through an earpiece.
The first device isn’t expected for at least another year, but there’s reportedly an obstacle between the vision and reality OpenAI’s audio models aren’t good enough yet. Currently, the Information reports, OpenAI researchers believe that their audio models aren’t as fast or accurate at answering user questions. To rectify this, OpenAI has reportedly “unified several engineering, product and research teams around the goal of improving audio models for its future devices.”
The Information reported that this push is headed by Kundan Kumar, a researcher previously at Character.ai. Other key figures include product research lead Ben Newhouse and ChatGPT product manager Jackie Shannon. Their efforts have produced a new architecture for audio models that the Information’s sources say are more natural, emotive, and accurate. A new audio model built on this architecture is expected to be released in Q1 of 2026.
The other challenge to OpenAI’s audio ambitions? Their current customers barely use ChatGPT’s existing audio features, according to a former OpenAI employee that the Information spoke with. For OpenAI’s devices to catch on, the company needs to train the AI-using public to embrace audio as an operating system. – Inc./Tribune News Service
