New report details OpenAI's screen-free AI home companion

Bloomberg journalist Mark Gurman, in reporting syndicated by the Los Angeles Times, says OpenAI is developing its first consumer device as a portable, screen-free AI speaker for the home. The product remains under development and has not been announced by OpenAI.

According to the report, the device would combine a camera and other environmental sensors with a rechargeable battery, allowing it to move between rooms rather than remain fixed like a conventional smart speaker. It is also said to include mechanical elements that can move, with OpenAI aiming to give the product a more expressive, companion-like presence. Reported functions include answering questions, responding to messages, playing media and helping control connected-home devices.

The same account says OpenAI is targeting a reveal in 2026 and a release in 2027, while warning that the schedule and design could change. An OpenAI spokesperson declined to comment on the unannounced product. Those details should therefore be treated as reported development plans, not confirmed specifications.

OpenAI has publicly confirmed the broader hardware program. Its July 2025 letter with Jony Ive says the io Products team merged with OpenAI, while Ive and LoveFrom retained major design responsibilities. That announcement confirms the partnership and product-development effort, but it does not identify a speaker, camera, sensor package or launch date.

OpenAI also introduced GPT-Live on July 8, 2026. The company describes it as a full-duplex voice system that can listen and speak at the same time, maintain a more natural conversational flow and delegate difficult tasks to a frontier model. Bloomberg's report connects a more advanced version of that voice technology to the planned device; OpenAI's GPT-Live announcement itself does not make that hardware connection.

If the reported design reaches market, its defining technical challenge will be persistent context rather than speaker hardware. A useful ambient assistant would need to coordinate audio, vision, memory, permissions and cloud tools without turning constant sensing into uncontrolled data collection.

For AI and data teams, the important open questions are:

• •Processing boundary: which audio, video and personal context would stay on-device, and which would be sent to OpenAI?
• •Permission design: how would users inspect, revoke and separate access to messages, email, cameras and smart-home controls?
• •Safety and reliability: what happens when a proactive assistant misreads context or acts on an incorrect inference?

Until OpenAI publishes product documentation, developers should treat the reported capabilities as directional signals about multimodal, voice-first agents rather than as a stable platform roadmap.