Apple is preparing to open CarPlay to third-party voice-controlled AI chatbots, including ChatGPT, according to people familiar with the company’s plans. The move, expected in the coming months, would give drivers a more capable conversational assistant when Siri cannot handle complex queries behind the wheel. But federal safety research and new academic work on large language models in vehicles suggest that “hands-free” does not mean “distraction-free,” raising hard questions about whether smarter voice AI actually makes driving safer.
What is verified so far
The strongest confirmed development is Apple’s plan to let outside AI chatbots operate within CarPlay through voice control. People described as familiar with the initiative told Bloomberg reporters that the feature is slated to arrive in “coming months,” though Apple has not publicly committed to a specific release date or version of iOS. If implemented, this would be a notable shift from CarPlay’s current design, which keeps tight control over how drivers interact with third-party apps.
Today, Apple’s own technical materials for car integrations describe a limited set of supported app types, including conversational experiences that work primarily through voice. Messaging, VoIP, and certain audio apps can appear in CarPlay, but they are funneled through SiriKit so that Siri mediates most interactions rather than letting each app define its own interface. This architecture is meant to keep the experience predictable and to minimize on-screen complexity during driving.
The consumer-facing iPhone User Guide reinforces that constraint by explaining that compatible third-party apps in CarPlay are controlled through Siri. Drivers can ask Siri to send messages, start calls, or play media, but they are not encouraged to browse arbitrary app interfaces. Allowing a chatbot such as ChatGPT to converse directly with the driver would loosen this control, potentially enabling more open-ended exchanges than Siri typically handles today.
On privacy, Apple has already outlined how ChatGPT fits into its broader “Apple Intelligence” strategy on phones, tablets, and Macs. In that context, the company says that when users invoke ChatGPT through Siri or system features, Apple proxies requests so that IP addresses are hidden and OpenAI does not store those queries by default. While Apple has not explicitly described the in-car architecture, any CarPlay integration would likely rely on the same underlying approach, since it is implemented at the operating system level.
The safety research is more sobering. A U.S. National Highway Traffic Safety Administration report on in-vehicle voice systems concluded that voice interactions reduce the need to look away from the road but do not remove cognitive distraction from secondary tasks. Drivers can keep their hands on the wheel and eyes forward while still devoting substantial mental resources to conversations, especially when tasks are complex or prolonged. The implication is that a more capable chatbot may not automatically translate into a safer driving experience.
Complementing that federal work, an academic team recently released an arXiv preprint that tested a large language model-powered conversational agent during actual on-road driving. The researchers compared interactions with the LLM-based assistant to simpler, hands-free baselines, measuring both visual attention and cognitive workload. Their study offers one of the first empirical looks at how a system resembling ChatGPT’s conversational mode might affect drivers relative to traditional voice assistants.
What remains uncertain
Even with these pieces in place, several critical gaps remain between what is known and what drivers would reasonably want to understand before relying on ChatGPT through CarPlay.
First, there is no primary crash or incident database that isolates collisions involving LLM-based chatbots in vehicles. The NHTSA voice-interface report speaks broadly about spoken interactions with in-car systems, not about open-ended AI dialogue. The arXiv study starts to address that gap but has not yet been peer-reviewed, and its specific findings about differences in cognitive load have not been replicated across multiple groups or driving environments.
Second, OpenAI has not issued a detailed public roadmap or safety protocol for in-vehicle use of ChatGPT. The Bloomberg account of Apple’s plans rests on anonymous sources and does not include formal statements from either company about how they will manage attention-sensitive situations. It is unclear, for example, whether the chatbot would automatically shorten responses when the vehicle is moving, pause when navigation alerts are issued, or provide any explicit cues to help the driver disengage quickly if road conditions change.
Third, automakers have not shared real-world testing data that specifically covers third-party AI chatbots running inside CarPlay. Apple’s developer rules describe which app categories are allowed and how they may present information, but they do not include empirical metrics about how conversational AI affects lane keeping, reaction times, or situational awareness. Without such data, claims that ChatGPT is safer or more distracting than Siri remain hypotheses rather than conclusions.
Fourth, the timing remains fluid. The Bloomberg report’s “coming months” phrasing is intentionally broad, and Apple has a history of adjusting software roadmaps without much public explanation. The feature could debut in a major iOS release, slip into a later point update, or be postponed if internal testing reveals new safety or reliability concerns. For now, the timeline should be treated as a sign that work is underway, not as a firm launch window.
Finally, there is little visibility into how regulators might respond once such integrations are widely available. NHTSA has shown interest in advanced driver assistance and infotainment systems, but there is no dedicated federal framework yet for evaluating or certifying LLM-based in-car chatbots. That regulatory uncertainty could shape how aggressively Apple, OpenAI, and automakers roll out or limit these features.
How to read the evidence
The available evidence divides into two broad categories, and keeping them distinct helps clarify what can and cannot be said about ChatGPT in CarPlay.
On one side is primary safety research. The NHTSA report on voice systems is a government-backed study with established methods, and its central conclusion (that voice control reduces manual and visual distraction but not cognitive load) is consistent with years of driving-safety literature. The arXiv LLM study is narrower but directly relevant, since it examines a conversational agent that behaves more like ChatGPT than like legacy command-and-control systems. Its on-road design and comparison to hands-free baselines make it an important early data point, even as readers should remember that preprints have not gone through formal peer review.
On the other side is contextual and business reporting. Bloomberg’s coverage provides the clearest picture so far of Apple’s product direction, but it is based on unnamed sources and omits technical specifics such as bandwidth limits, interaction caps, or UI safeguards. Apple’s developer documentation and consumer guides tell us how CarPlay works today and what categories of apps are allowed, but they do not yet describe LLM-specific behavior. Apple’s Apple Intelligence announcement fills in the privacy model for ChatGPT access at the OS level, but it stops short of addressing in-car ergonomics or safety trade-offs.
Readers should also be aware of the information environment around this story. Business-focused outlets such as Bloomberg’s terminal emphasize product timelines and competitive positioning, while safety researchers foreground crash risk and cognitive load. Neither perspective is complete on its own. A cautious interpretation treats Apple’s reported plans as credible but not guaranteed, and the current safety evidence as suggestive but far from definitive about LLM chatbots in cars.
For drivers, the practical takeaway is that a smarter voice assistant can reduce the temptation to pick up the phone but cannot make the mental demands of conversation disappear. Until more targeted research and real-world data are available, it is reasonable to view ChatGPT on CarPlay as an incremental convenience rather than a proven safety upgrade, and to use any in-car conversational AI with the same restraint that safety agencies already recommend for hands-free calls and messages.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
