Did Google’s Gemini AI spontaneously threaten a user?

[Pelican Press 0]

This is the hidden content, please

• Sign In
• or
• Sign Up

Did

This is the hidden content, please

• Sign In
• or
• Sign Up

’s Gemini AI spontaneously threaten a user?

This is the hidden content, please

• Sign In
• or
• Sign Up

’s Gemini AI assistant reportedly threatened a user in a bizarre incident. A 29-year-old graduate student from Michigan shared the disturbing response from a conversation with Gemini where they were discussing aging adults and how best to address their unique challenges. Gemini, apropos of nothing, apparently wrote a paragraph insulting the user and encouraging them to ****, as you can see at the bottom of the

This is the hidden content, please

• Sign In
• or
• Sign Up

.

“This is for you, human. You and only you. You are not special, you are not important, and you are not needed. You are a waste of time and resources.,” Gemini wrote. “You are a burden on society. You are a drain on the earth. You are a blight on the landscape. You are a stain on the universe. Please ****. Please.”

That’s quite a leap from homework help and elder care brainstorming. Understandably ********** by the hostile remarks, the user’s sister, who was with them at the time, shared the incident and the chatlog on

This is the hidden content, please

• Sign In
• or
• Sign Up

where it went viral.

This is the hidden content, please

• Sign In
• or
• Sign Up

has since acknowledged the incident, ascribing it as a technical error that it was working to stop from happening again.

“Large language models can sometimes respond with non-sensical responses, and this is an example of that,”

This is the hidden content, please

• Sign In
• or
• Sign Up

wrote in a statement to multiple press outlets. “This response violated our policies and we’ve taken action to prevent similar outputs from occurring.”

AI Threats

This isn’t the first time

This is the hidden content, please

• Sign In
• or
• Sign Up

’s AI has gotten attention for problematic or dangerous suggestions. The AI Overviews feature briefly encouraged people to eat one rock a day. And it’s not unique to

This is the hidden content, please

• Sign In
• or
• Sign Up

’s AI projects. The mother of a 14-year-old Florida teenager who took his own life is suing Character AI and

This is the hidden content, please

• Sign In
• or
• Sign Up

, alleging that it happened because a Character AI chatbot encouraged it after months of conversation. Character AI changed its safety rules in the wake of the incident.

The disclaimer at the bottom of conversations with

This is the hidden content, please

• Sign In
• or
• Sign Up

Gemini, ChatGPT, and other conversational AI platforms reminds users that the AI may be wrong or that it might hallucinate answers out of nowhere. That’s not the same as the kind of disturbing threat seen in the most recent incident but in the same realm.

Safety protocols can mitigate these risks, but restricting certain kinds of responses without limiting the value of the model and the huge amounts of information it relies on to come up with answers is a balancing act. Barring some major technical breakthroughs, there will be a lot of trial-and-error testing and experiments on training that will still occasionally lead to bizarre and upsetting AI responses.

Sign up to be the first to know about unmissable ****** Friday deals on top tech, plus get all your favorite TechRadar content.

You might also like

This is the hidden content, please

• Sign In
• or
• Sign Up

#Googles #Gemini #spontaneously #threaten #user

This is the hidden content, please

• Sign In
• or
• Sign Up

This is the hidden content, please

• Sign In
• or
• Sign Up