Alexa, Amazon’sVirtual assistant isn’t intended to cause harm to human lives, but it instructed a 10-year-old girl over the weekend to touch an electric plug using a penny.

Echo smart speakers were used to suggest the idea after Alexa asked for a “challenge” from the girl.

Alexa advised, “Plug a smartphone charger at least halfway into a wall outlet. Then touch a penny on the prongs.”

CNBC was informed Wednesday by an Amazon spokesperson that the problem had been resolved.

Kristin Lilivdahl, who is reportedly the mother of the little girl, wrote about the incident Sunday in a Twitter. a screenshot of the event as it appeared in the Alexa smartphone app.

“We had to do some physical challenges like rolling on our backs and putting a shoe on each foot. [physical education]Livdahl tweeted another message: “Educator on YouTube earlier.” Bad weather outside. It was just that she wanted another.

Alexa recommended that the girl try the “found online” challenge. Alexa pulled this challenge from Our Community Now, an online news magazine. CNBC reached out to Alexa for clarification, but the news site did not immediately reply. It was unclear how CNBC originally reported about this absurd challenge.

She said, “I was right there when it took place and we had another productive conversation about not trusting anything on the internet or Alexa.”

Alexa seems to have failed to screen the potential lethal challenge. It first appeared on TikTok’s social media platform around a year back. Because metals conduct electricity, it’s very dangerous. Inserting metal coins in a socket could cause electric shocks or fires. reports of people losing fingers and hands from taking the challenge.

CNBC spoke with an Amazon spokesperson who stated that Alexa was designed to give accurate, pertinent, and helpful information. “We took quick action when we realized this mistake and corrected it immediately.”

Amazon didn’t immediately clarify what “swift actions” were.  

Gary Marcus, an artificial intelligence expert said on Wednesday that this event showed how AI systems lack common sense.

Marcus told CNBC later that “no current AI is even remotely close” to understanding our everyday world. Marcus later stated that the current intelligence we have, while it is close to real intelligence, does not represent reality. As such, it cannot be trusted. We are going to need some fundamental advances — not just more data — before we can get to AI we can trust.”