Tony Fadell Says AI Gadgets Need Trust, Privacy and Clear Purpose

Tony Fadell believes the first generation of dedicated AI gadgets failed for a basic reason: they showcased new technology without solving a clear consumer problem. Speaking at MIT Future Fest, the iPod co-creator and Nest founder pointed to early AI devices such as the Rabbit R1, Humane AI Pin and Limitless pendant as examples of products that struggled to establish a lasting role in everyday life.

Fadell said companies behind some early AI devices approached him for advice, but he declined to get involved. His criticism centres less on the underlying technology than on product-market fit. Consumers, he argued, need to understand what problem a device solves before its technical capabilities matter.

Personal AI assistants face a trust problem

That challenge becomes more complicated as companies move from simple chatbots toward AI assistants that can act on a user’s behalf. Fadell noted that relatively few people have ever employed a human personal assistant, meaning many consumers have little experience deciding what responsibilities to delegate or how much access to provide.

Trust with a human assistant also develops gradually. Fadell said it took him years to learn how to work effectively with one before he felt comfortable sharing sensitive information or letting the assistant handle tasks involving meetings and financial matters.

AI agents raise similar questions at much larger scale because useful assistants may need access to email, payments, contacts, location and other private information. Recent security issues involving Meta’s Muse illustrate the stakes. A security vulnerability reported after Muse launched raised concerns about potential access to sensitive user data, while 404 Media reported that Meta rushed to address separate security flaws before release.

Fadell sees on-device AI as critical

Fadell argues that successful personal AI assistants will eventually need to perform much of their work directly on users’ devices rather than continually transmitting sensitive information to cloud infrastructure. In his view, improvements in local processors and computing efficiency will make that approach increasingly practical.

He singled out Apple as particularly well positioned because it already controls widely used hardware, custom chips and operating systems while maintaining a strong consumer privacy position. Apple’s weakness, he argued, is its relative position in advanced AI models, an area where the company has increasingly relied on outside technology.

Fadell has not shied away from criticising his former employer. Earlier this year, he criticised Apple’s handling of AI and its abandoned car project, while suggesting the company’s hardware and privacy strengths could still give it advantages in new AI devices.

Why AI companies are building their own hardware

Fadell also explained why companies such as Meta and OpenAI are investing in dedicated AI hardware. Unlike Apple, they do not control billions of consumer devices with built-in cameras, microphones, location sensors and other hardware that an AI assistant could use.

A separate gadget can give an AI company direct access to those sensors without depending entirely on permissions and interfaces controlled by smartphone platforms. But Fadell’s broader argument is that collecting more data or adding more hardware will not solve the central product challenge on its own.

The next generation of AI gadgets, in his view, will need to demonstrate a specific everyday purpose while earning enough trust for users to hand over meaningful responsibilities. For startups in particular, getting that combination wrong can be costly because they may have far less room than an established hardware company to recover from a failed first product.