GM is turning OnStar into something more ambitious than a breakdown service with a button. The company today announced the OnStar AI Assistant, a conversational in-vehicle intelligence layer that takes natural language commands, answers vehicle-specific questions, and hands off to a live Advisor when the situation demands a human. The first production deployment arrives in the 2027 Cadillac LYRIQ by over-the-air update in the coming months.
The wake word is “Hey OnStar.” From there, the assistant interprets intent rather than scripted commands. Say the cabin is warm and the temperature adjusts. Ask whether there is enough charge to reach a destination and the assistant checks available range, then maps charging stops if needed. Ask what a dashboard warning light means and it pulls an explanation specific to that vehicle rather than a generic lookup. That last part matters: a general-purpose voice assistant can tell you what a tire-pressure symbol typically means; the OnStar AI Assistant is trained on GM’s own vehicle documentation and knows the exact model in front of it.
Google Gemini powers the underlying AI, extending a partnership that dates to 2019, when GM first began embedding Google software across its lineup. GM began rolling Gemini out to roughly 4 million vehicles from model year 2022 and newer earlier this year, replacing Google Assistant. The OnStar AI Assistant builds on top of that infrastructure, adding vehicle control, real-time sensor access, and the ability to escalate to a human Advisor within the same session.
The competitive landscape has moved quickly. Volkswagen integrated OpenAI’s ChatGPT into its IDA voice assistant across a wide range of models in 2024, and Mercedes-Benz has been testing ChatGPT in more than 900,000 vehicles in the United States through the “Hey Mercedes” system. Stellantis extended its own arrangement with Mistral AI to develop in-car assistance tools. GM’s answer is to lean on the 30 years of vehicle and customer data OnStar has accumulated across more than 12 million subscribers in 22 markets, arguing that depth of vehicle-specific knowledge is harder to replicate than raw language model capability.
GM has not announced subscription pricing for OnStar AI Assistant.
The New OnStar App
Starting in 2027, a new OnStar mobile app will replace the current OnStar Guardian app and the separate Buick, Cadillac, Chevrolet, and GMC brand apps. Four apps become one. The consolidated app extends several features that previously required being in the vehicle.
The Connected Cameras function is the most operationally distinct addition. When the vehicle detects a trigger event, such as an aggressive horn honk, onboard cameras capture the incident, encrypt the recording, and push it to the OnStar app with an alert. GM states that it cannot access Connected Camera recordings; drivers control what is stored and for how long. The feature is opt-in, and users can manage camera, microphone, location, and AI permissions through a dedicated settings panel with clear status indicators.
OnStar Family ties multiple vehicles and users into a shared interface, letting family members share destinations, coordinate pickups, monitor other vehicles in the household, and receive safety or community alerts. A route-awareness layer combines vehicle sensor data, weather, and road conditions to surface proactive guidance before a trip turns difficult.
Fleet Intelligence
The fleet side of this announcement largely extends what GM introduced with OnStar Fleet Intelligence in September: a unified dashboard with AI-driven analysis, natural language queries, and visibility into vehicle health, maintenance schedules, and fuel and energy consumption. The OnStar Fleet AI Assistant lets fleet managers ask operational questions in plain language and receive trend analysis rather than raw data exports. The same privacy architecture that governs consumer features applies to the fleet platform.
Privacy Architecture
GM built opt-in activation into every feature. Nothing runs unless a customer enables it. The settings layer shows which data each feature uses, why it is needed, when it is active, and how it is handled. Recordings are time-limited and encrypted. The framework is worth noting because connected-camera and always-on microphone features have met real consumer resistance in other products; GM is addressing that friction before it becomes a barrier to adoption.
OnStar’s 30-year history is a genuine asset here. The service has accumulated driver-assistance credibility that a software company launching a car feature from scratch cannot claim. Whether the AI layer lives up to that credibility will show in the first months of real-world LYRIQ use. The LYRIQ gets it first; the rest of the GM lineup follows on a schedule the company has not yet specified.
Source: General Motors. Images courtesy of General Motors.
