A companion for you, a “body” for ChatGPT

TL;DR: OpenAI is reportedly working on a “humanlike” AI companion device designed to feel alive. It’s hardware that people will likely have to pay for, from a company whose flagship product is mostly used for free—but an AI companion gadget also comes with a lot of social (and, increasingly, legal) baggage.

What happened: You might mistake OpenAI’s upcoming device for an AI-powered speaker at first—it has no screen (cue the Jony Ive jokes), uses a more advanced version of ChatGPT’s Voice mode to interact, and can play music.

But, per Bloomberg, the device’s key feature is its personality, which grows more distinctive and proactive as it learns who you are. Insiders say it’s a “new type of home computer for the AI era,” with “mechanical elements that can move on their own.” The pitch appears to be to turn a chatbot people already treat as a confidant into a physical object that can live on your kitchen counter (hello, Alexa) or go wherever you go—and could cost hundreds of dollars, according to the Information.

Details about the device, which OpenAI reportedly aims to launch in 2027, also conveniently come just days after Apple sued the company for allegedly stealing hardware trade secrets (though sources say the product isn’t similar to anything Apple sells).

The loneliness-tech era: About 16% of US adults have used AI for companionship as of last year, per a poll—while almost three-quarters of US teens have tried out AI companions.

But that doesn’t mean everyone’s welcoming AI pals with open arms. Mental health experts criticize them as a Band-Aid for loneliness; others even argue that AI companions could worsen the downslide in birth rates. And then there was the public’s very visceral, spray-painted rejection of Friend, a wearable AI pendant sold as an always-listening buddy that could do things like binge-watch TV shows with you.

The regulatory reckoning: AI companions are already getting plenty of political and legal pushback. Several chatbot makers, including OpenAI, have been hit with lawsuits alleging their products are addictive, foster unhealthy emotional dependency, and even encourage self-harm. Momentum to require guardrails, or outright ban AI companions for minors, is gaining steam. Case in point: Rules that went into effect today in China aim to curb people’s emotional attachment to AI.

Bottom line: OpenAI’s attempt to monetize our impulse for companionship comes at a time when the company could use another ChatGPT-level win—but it may also be reaching for a market before its time, one currently beset by regulatory hurdles, social backlash, and mounting lawsuits. —WK

Also at OpenAI…

• The company says it did actually respond to Apple’s concerns about it stealing trade secrets… but talks collapsed after an outside attorney Apple hired made a really uncomfortable blunder.