General Motors General Motors had its Q1 2026 earnings call last week and CEO Mary Barra dropped an interesting tidbit in her opening minutes: "Today, nearly 90% of the code written by our autonomy team is generated by AI." That autonomy team is currently working on the next-generation, "eyes-off, hands-off" rendition of Super Cruise set to debut on the Cadillac Escalade IQ in 2028. This being an earnings call with an audience primarily made up of money guys, Barra rattled off the "90%" stat as a badge of honor: "The way we're building this technology is a reflection of how seriously we're embracing AI across the enterprise," she said. Look, as someone who did indeed take first-year college computer science, I feel like enough of software creation is rote and mechanical enough such that having AI write portions of code isn't as problematic as, say, having AI write this article you're reading right now. But still, 90% feels like a lot. For reference, a study by Science.org estimates that 29% of code produced in the U.S. by 2025 was "produced with substantial AI support." High-stakes code General Motors What's more, Barra isn't talking about having AI produce in-car weather apps or whatever. This is Super Cruise, the autonomous driving tech that's supposed to let Chevys, Cadillacs, GMCs, and Buicks practically "drive themselves" and keep occupants, pedestrians, and other vehicles on the road safe. And given this is GM and not some boutique EV brand that'll only sell 150 cars in SoCal, it's not an exaggeration to say that any errors here could easily become issues of public safety. In any case, Barra says next-gen Super Cruise will be deployed "on both ICE vehicles and EVs and scale across multiple brands and price points." GM is apparently able to digitally stress test the technology by "simulating roughly 100 years of human driving every single day," in addition to recently beginning to do real-world road-testing in California and Michigan. For all our sakes, I genuinely hope the next version of Super Cruise works as advertised, GM's autonomy team is indeed having qualified humans review all of the AI code, and that we can eventually look back on this story and be like, "That turned out fine, actually." But given corporate America's long, storied history of compromising the well-being of the public in the name of hype-driven cost-cutting and the fact that ChatGPT once tried to convince me that the Honda S2000 does not have a cabin air filter (hit me up on Instagram if you'd like a screenshot of this), that optimism should be taken very, very cautiously.
+25 others
