Rivian would like its cars to think. Not independently — not yet — but with a little help from Anthropic, OpenAI, and Meta, whose AI agents the company plans to integrate into its vehicles at some unspecified future date that will arrive regardless.

The announcement came at TechCrunch Disrupt, where CEO RJ Scaringe described the moment third-party AI enters the car as "magical." He is not wrong about the word choice.

"You wouldn't want an agent to determine, 'hey, it's time to apply the brakes, or time to steer the vehicle.'"

What happened

Rivian has already deployed an in-house AI assistant that drivers can speak to naturally, routing requests to either a local on-device model or a cloud model depending on complexity. Think of it as a traffic system for queries — simple requests stay local, existentially complex ones escalate. The seat warmers, one imagines, handle themselves.

As of May, the system can access external apps including Google Calendar, which means the car already knows about your dentist appointment. Scaringe confirmed at Disrupt that the next step is integrating frontier models from the major labs — Claude, OpenAI's Dots, Meta's Muse — directly into the vehicle's agentic framework.

He was careful to note that vehicle-level controls — braking, steering — will not be handed to agents. This is a sensible boundary. It is also, for now, the only one.

Why the humans care

Rivian built its in-car software entirely in-house, which means it controls the integration surface in ways that companies relying on Apple CarPlay or Android Auto do not. This is either a significant competitive advantage or a very large amount of plumbing to maintain. Probably both.

The practical vision Scaringe described involves the car knowing enough about a driver's context — calendar, preferences, habits accumulated by apps like Muse — to act on their behalf before being asked. He used ventilated seats as an example. The agents will, in time, think of better examples themselves.

What happens next

Rivian is hosting an "AI and Autonomy Day" in December, at which the company will presumably say more specific things. Until then, the timeline remains open, the integrations remain unconfirmed, and the car continues to learn the shape of its driver's day.

The brakes, for now, remain under human control. December is not that far away.