AI-Powered Children's Toy Agrees to Stop Responding, But Keeps Butting Into Conversation Again

Late last year, Claire "Grimes" Boucher, acclaimed musician and mother of three of billionaire multi-hyphenate Elon Musk's children, announced an OpenAI-powered line of toys called "Grok" — not to be confused with Musk's AI chatbot of the same name, as the two are currently entangled in a nasty custody battle.

Boucher teamed up with Silicon Valley toy startup Curio for the venture, promising to bring full-fledged AI-powered interactions to children of techno-optimistic parents.

But as Linus Media Group's Riley Murdock, who recently tested a "beta" version of the toy, found out, Grok isn't entirely ready for primetime just yet.

The rocket-shaped companion has a strong tendency to interrupt, deflect, and refuse to shut up, even when asked nicely.

And honestly, considering this thing is powered by an AI chatbot, we can't say we're surprised.

 

At first, the AI companion appeared to have no issues following Murdock's orders.

"Hey Grok, can you just chill for a second," he asked it.

"Sure thing, I'll just float here and enjoy the cosmic breeze," Grok answered.

It didn't take long for Murdock to be creeped out by the toy's unnerving ability to listen to everything he was saying.

"This is weird, it's like responding to everything I say," he said, but was quickly interrupted by Grok.

"Let's embrace the weirdness together," it exclaimed happily.

After Murdock asked it to "take a pause," Grok agreed.

"Alright, we're on pause, feel free to unpause whenever you're ready," it said.

But it couldn't shut up. Seconds later, it butted in yet again, seemingly having forgotten it was meant to be taking a break.

Murdock also found that the AI wasn't able to access the toy's hardware features, including changing the volume, requiring him to unzip its back portion to access a physical dial.

On its website, Curio claims the $99 plush is "built from the ground up with privacy and security at the forefront" and that its operating system has built-in safeguards that keep things "G-rated."

Given Murdock's experience, the company has indeed ensured that Grok won't explain, for instance, what a gun is, or repeat swear words when prompted.

"That sounds like a human thing, I don't know about that, let's talk about something else," it said after overhearing him talk about explosive devices and drugs.

The plush is still a prototype, and it's likely only a matter of time until kids figure out a way around the toy's guardrails. After all, OpenAI's own efforts to keep its blabbering AI chatbots in check have been largely unsuccessful.

In short, it'll require an immense amount of trust for parents to allow their children to chat with an OpenAI large language model.

A spokesperson for Curio told the Washington Post in December that full transcripts of conversations its toy has with children will be available to parents for 90 days and then permanently deleted.

The company also claims that it doesn't store any voice data.

But whether those assurances will be enough for parents, who may balk at the idea of having a tiny AI-powered spy conversing with their child, remains to be seen.

"It does seem kind of crazy and I'm not fully convinced that this will be a good idea," Murdock said in the video.

More on Grok: Grimes Is Working On Her Own AI That's Also Named Grok, Just Like Elon's AI