General Motors will launch a proprietary in-vehicle artificial intelligence assistant across its vehicle lineup later in 2026 [1].
The move signals a shift toward deeper hardware and software integration. By developing its own system, GM aims to provide capabilities that third-party software cannot currently offer, specifically regarding real-time vehicle telemetry and predictive maintenance.
The new AI assistant will be integrated directly with OnStar, the company's subscription-based concierge and safety service [1, 2]. This integration allows the AI to access deeper vehicle data, enabling the system to monitor the health of the car and alert drivers to potential mechanical issues before they become critical [3, 4].
Previously, GM deployed Google's Gemini system within its vehicles. However, the company determined that a proprietary solution was necessary to achieve the level of vehicle-specific integration required for its long-term goals [3, 5]. The proprietary system is designed to handle complex queries that require a direct link to the vehicle's internal computer systems, a level of access typically restricted for external developers.
This rollout will affect both U.S. and global markets [1, 3]. The company expects the system to improve the user experience by streamlining how drivers interact with vehicle settings and safety features. By controlling the AI stack, GM can update the system over-the-air to add new telemetry-based features without relying on external software partner timelines [2, 4].
The launch is scheduled for later in 2026 [1].
“GM aims to provide capabilities that third-party software cannot currently offer.”
This transition reflects a broader trend in the automotive industry where manufacturers are attempting to reclaim the 'digital cockpit' from tech giants like Google and Apple. By owning the AI layer, GM can monetize proprietary data and create a more seamless ecosystem between the vehicle's mechanical health and the user interface, potentially increasing the value of its OnStar subscriptions.

