Speaking at the inaugural MIT Future Fest, Tony Fadell highlighted a slide featuring three discontinued AI devices: the Rabbit R1, the Humane AI pin, and the Limitless pendant. Fadell, widely recognized for his role in creating the iPod and iPhone and for founding Nest, declined offers to consult for these startups, noting that they prioritized novel technology over genuine consumer needs.
According to Fadell, these products attempted to replicate the utility of a personal assistant despite only a tiny fraction of the global population having ever hired human help. He emphasized that trust does not come easily; it took him years to feel comfortable allowing a human assistant access to sensitive information like banking details. Transferring that level of trust to AI agents remains a significant hurdle.
The challenges around security were illustrated by Meta’s recent launch of Muse, its AI assistant. Reports emerged shortly after release revealing serious vulnerabilities, including internal awareness of security flaws that prompted an urgent, overtime effort by multiple teams to patch issues before the product went public.
Fadell asserted that safety and security will be decisive factors for any AI tool that users allow to manage their private data. He suggested that currently, only Apple—and possibly one other company—has the infrastructure to handle this responsibility effectively. While Apple leads in hardware and has built significant consumer goodwill regarding privacy through features like Face ID, it relies on Google’s Gemini models for its new Siri AI capabilities.
Looking ahead, Fadell predicted that successful AI agents will likely operate entirely on-device to ensure privacy and maintain lightweight performance. He dismissed the notion that cloud-based data centers would dominate, citing the immense computing power now available in battery-operated devices.
He also theorized that companies like Meta and OpenAI are developing standalone hardware because they lack access to the vast array of sensors found in smartphones. Without the billions of connected devices that Apple possesses, these tech giants are building external gadgets to capture video, audio, and location data via Bluetooth or Wi-Fi.
Despite his past success with Apple, Fadell cautioned startups about the difficulty of achieving product-market fit. Unlike established giants such as Apple, which can afford to iterate on products like the Vision Pro, startups typically have only one opportunity to succeed before financial constraints force them out of business.
The security vulnerabilities in Meta’s Muse are terrifying. How can we trust these agents with our banking info when basic patching is this chaotic? Trust has to be earned first.
I’m skeptical we’ll see purely on-device AI soon. The battery drain and thermal issues seem unsolvable with current silicon. Cloud processing is probably sticking around longer than he thinks.
Spot on about the Rabbit R1 and Humane. They looked cool but solved nothing. Apple really does understand hardware integration better than anyone else right now.