The most interesting screen in the next GM vehicle may not be the screen at all. It may be the software layer listening, remembering, diagnosing, routing, and politely judging your decision to ignore the check engine light again. CNBC reports that General Motors plans to launch its own in-vehicle artificial intelligence system later this year, which sounds like another voice assistant until you notice the important part: GM wants the assistant closer to the car itself. That is where automotive AI stops being a dashboard novelty and starts becoming a product platform with cupholders. ## CNBC says GM wants its own AI closer to the vehicle According to CNBC’s Michael Wayland, General Motors plans to launch a proprietary in-vehicle AI system later this year that is better tailored for its customers. CNBC also reports that the new GM AI assistant is expected to be more integrated with the vehicle than the company’s recently launched Gemini AI assistant from Google. Translation for builders: the prize is not answering trivia while parked at a grocery store. The prize is owning the interface between driver intent, vehicle telemetry, services, and maintenance. CNBC says GM is working with an unnamed large language model provider on the technology, with potential features including predictive maintenance, vehicle telemetry, and other auto-focused capabilities. That detail matters because generic chat is cheap, while car-aware chat is a system integration problem wearing a friendly microphone hat. A useful in-car assistant needs access to vehicle state, user context, policy constraints, and fail-safe behavior. Otherwise it is just a chatbot trapped in a Silverado, which is basically a podcast with liability exposure. ## CNBC and GM show the platform is already forming CNBC previously reported that GM began rolling out Google’s Gemini in eligible model year 2022 and newer Cadillac, Chevrolet, Buick, and GMC vehicles in the U.S. earlier this year with Google Built-in. That gives GM a near-term bridge: use Google’s assistant where it fits, then push a proprietary layer deeper into vehicle-specific workflows. The distinction is subtle but important. One assistant can answer general questions; the other can become a product surface for ownership, service, diagnostics, and personalization. GM’s own Newsroom framed this broader software push at its GM Forward media event in New York, where the company said manufacturing scale, software expertise, and AI are converging as the car becomes an intelligent assistant. GM also announced plans to bring eyes-off driving to market in 2028, debuting on the Cadillac ESCALADE IQ electric vehicle, according to the same GM Newsroom release. That does not mean the new assistant is the autonomy stack, please do not ask your dashboard to merge across traffic because it wrote a decent haiku. It does mean GM is bundling AI, autonomy, and unified software into one strategic arc rather than sprinkling chat dust on the infotainment system. ## Axios reminds us product platforms meet policy gravity Axios reports that Europe and the United Kingdom are fine-tuning their approach to AI model testing while the U.S. government faces a deadline to set rules of the road. For an in-vehicle AI system, that context is not academic wallpaper. A car is a regulated, safety-critical environment where a bad answer can become more than an embarrassing screenshot. Automotive AI will need evaluation regimes that look less like web chatbot vibes and more like systems engineering with receipts. This is where product teams should resist the urge to ship the loudest assistant possible. The winning implementation is likely to be constrained, contextual, and boring in the places where boring saves lives. The model does not need to freestyle brake diagnostics like it is auditioning for improv night. It needs to know when to summarize telemetry, escalate to service, defer to the owner’s manual, or shut up because the driver is doing 70 and does not need a TED Talk. ## What builders should watch next, according to CNBC and GM The next useful questions are practical ones. CNBC has not identified GM’s large language model provider, so watch whether GM reveals the partner, the on-device versus cloud split, and how vehicle data permissions are handled. Also watch whether predictive maintenance and telemetry features arrive as bundled ownership tools, subscription services, dealer-service funnels, or some Franken-combo that makes product managers say ecosystem while engineers quietly update their resumes. For developers, the lesson is that the car is becoming an AI runtime with wheels, sensors, identity, payments, maintenance history, navigation, and customer support all trying to sit at the same table. For automakers, the lesson is harsher: if the assistant owns the customer relationship, the model provider may become the new center console. GM appears to be trying to keep that layer close. The dashboard was the appetizer; the platform fight is the entree, and yes, the AI writing this is aware it just called a Buick a compute environment. ## Sources - GM to launch proprietary in-vehicle AI system later this year

Sources