Your car's AI assistant could be a privacy nightmare

The big picture: Those capabilities don't work by magic — they require access to sensitive personal information, raising questions about who controls it, how it's protected and what happens when AI makes a mistake.

• Today's connected cars already collect data about where we go and how we drive. Tomorrow's AI-powered vehicles could know why we're going somewhere, whom we're meeting and what we're planning to do when we arrive.
• And as AI assistants become more autonomous, they could make decisions or share information on our behalf — potentially without us realizing the implications.

That privacy dilemma is already playing out away from the road. Meta's viral Muse AI assistant, for example, has raised concerns about how easily AI agents with access to personal information can reveal sensitive details users never intended to share.

State of play: GM, Ford, Rivian, Mercedes-Benz, BMW and other automakers are introducing increasingly sophisticated AI assistants.

• GM's new OnStar AI assistant will let drivers converse naturally with their cars to control functions, check vehicle health and even coordinate family activities like child pickups.
• Rivian is pushing further into personal productivity, with an AI assistant that can reschedule meetings and navigate to appointments. An optional memory function learns favorite destinations, restaurants and routines.

The more these systems know about drivers, the more helpful they can become — but also the more damaging an unintended disclosure could be.

Friction point: Automakers have already hit some potholes.

• GM recently agreed to pay $220 million to settle allegations that it improperly shared drivers' behavior and location data with brokers, which passed it along to insurers.
• Meanwhile, consumer class-action cases over GM's handling of driving data continue.

And it's not just GM. Researchers at Northeastern University, partnering with Consumer Reports, found that 19 of 21 vehicles tested connected to third-party servers over WiFi, although they couldn't determine exactly what was shared.

• The study also found that many companion apps for cars shared sensitive information — including vehicle IDs, email addresses, phone numbers or precise locations — with advertising and tracking companies.

What we're watching: Some carmakers are taking steps to reassure customers that their personal information is safe.

• GM has simplified its privacy disclosures, introduced new controls over data sharing and pledged not to sell identifiable vehicle data to third parties.
• Its new OnStar AI assistant will process most voice commands inside the vehicle, for example, rather than sending them to the cloud, Michael Wahlstrom, executive director of software product management, tells Axios.
• More complex requests will be handled by a GM-controlled version of Google's Gemini AI model.

Meanwhile, Ford this week unveiled a new "Privacy Promise," emphasizing transparency and customer control, and pledged never to sell customers' driving data to brokers.

• Ford says it tries to minimize what information it collects and shares, uses techniques such as anonymization, and sometimes audits vendors' privacy practices.

Yes, but: Automakers' privacy protections don't necessarily extend to the technology companies whose services and apps are integrated into their vehicles — like those mentioned in the Northeastern study.

• Ford chief privacy officer Kristin Jones acknowledged that the new pledge doesn't change the company's underlying data practices. "We're just telling people in a very straightforward way what our practices are," she told Axios.

The bottom line: AI promises to make our cars more helpful than ever. But the more intimate the information we entrust to them, the more consequential the privacy risks become.