Rivian Prepares for Integration of Third-Party AI Agents with Safeguards

Rivian already has a voice assistant in its electric vehicles and describes an architecture that relies on either local models or cloud-based solutions, depending on the case. RJ Scaringe aims to go further with third-party agents, although no timeline has been announced, and emphasizes strict safety limits. More details will be provided at an event in December.

Exclusive Control of Agents and Complex Integration, According to Scaringe

RJ Scaringe explains that integrating AI agents into Rivian vehicles requires significant technical development and close collaboration with AI labs to enable this connection. He stresses the need for great caution, as this software is closely related to systems essential for vehicle safety. Functions such as braking or steering will not be entrusted to AI agents. No timeline has been communicated for this evolution, with Rivian referring to the upcoming AI & Autonomy Day event in December for more information.

Voice Assistant Already Deployed and Local or Cloud Model Selection

Rivian's AI assistant is already available, allowing drivers and passengers to interact with their vehicle via voice commands. Rivian states that it has designed an agentic framework capable of understanding user requests and directing them to the appropriate software function, for tasks such as adjusting the temperature, activating heated seats, or summarizing news. According to RJ Scaringe, the system can choose to use an embedded local model or, for more complex requests, send the query to the cloud, where different models are mobilized. Since May, the assistant can also access external applications like Google Calendar.

Potential Opening to External Agents and Market Context

During TechCrunch Disrupt, RJ Scaringe mentioned that Rivian vehicles could in the future run AI agents developed by other companies, such as Anthropic. He envisions an evolution where the use of agentic applications would be extended by connecting the car to the context that these services, such as Muse, already hold about their users. He gives the example of a driver activating the ventilated seats while asking for help organizing a meal, with the car then able to leverage the context of a previous conversation with the assistant on another device. This direction aligns with Rivian's software strategy, which has developed its embedded system without resorting to Apple CarPlay or Android Auto, has internally designed applications like navigation, and is working to integrate popular services like Apple Music. In the market, Tesla has already added this year a feature allowing interaction with SpaceX's Grok chatbot and its Grok Bot assistant.