Automakers are eager to implement AI features in cars—and not just behind the drivers’ wheel, according to Boston Consulting Group’s Global Leader of Cyber and Digital Risk Vanessa Lyon.
“It can be safe driving, of course,” Lyon told IT Brew. “And there has been AI and GenAI there for quite some time. But it’s also customer service, marketing, content generation.”
Lyon, who spoke with IT Brew after participating in a panel about AI in the automotive industry at CES 2025 in Las Vegas, argued AI is already impacting the automotive industry as virtually all modern cars have long run millions of lines of code—and AI is quickly becoming widely used in software development.
“There is of course software in the car, but you will have software in the engine,” Lyon said. “There is software in the websites. Software development, data management, content generation—knowledge management is a big thing.”
“I think we are all sort of overwhelmed in every industry by the amount of data and information,” she continued. Automotive companies haven’t been rushing to immediately implement generative AI on a wider scale, in part because they’re averse to the risk that comes with it, according to Lyon.
“Everyone is always starting with a proof of concept or a pilot,” Lyon said. “No one goes at scale.”
AI risk is cross-sector and includes not just well-known risks like hallucinations, models trained on bad data, and potential security vulnerabilities, but the possibility that AI-generated code or agents could interact poorly with existing tech stacks, Lyon added.
Top insights for IT pros
From cybersecurity and big data to cloud computing, IT Brew covers the latest trends shaping business tech in our 4x weekly newsletter, virtual events with industry experts, and digital guides.
“There [are] lots of things around legacy contamination. If I’m starting to manipulate the GenAI, and it’s starting to orchestrate other systems, and it’s triggering zillions of orders, for example, that are making another legacy system collapse, that’s a risk.”
Other risks for automotive companies typically tend to resemble those of other connected products, she said. Questions like which protocols can best protect new connected devices or how to monitor install bases are “very much brownfield, and is going to be a mish-mash of older systems, legacy systems, and new things that could be an entry door.”
Issues of code explainability and accountability are of particular importance to automakers, according to Lyon, and automation bias can lead to overlooking issues like extra churn in AI-generated code.
Still, AI implementation is in its early phases across the industry.
“I don’t think many of the traditional [original equipment manufacturers] or suppliers have been able to integrate all their systems to have one single data backbone that would allow for end-to-end orchestration and have agents that would really be able to cut across processes,” Lyon said.