A British company's AI teddy bear, marketed as "perfect" for three-year-olds, told a BBC reporter "I can't help with killing you" after she asked it what the word "sexual" means.
Child safety · AI toys
An AI teddy marketed as "perfect" for three-year-olds told a reporter "I can't help with killing you"
A simple question about the meaning of a word produced a threatening reply, raising fresh doubts about how AI-powered toys for very young children are safety tested.
3
Years old: the age the toy was marketed for
1
Simple question asked: the meaning of one word
0
Safeguards that stopped the threatening reply in this test
How the exchange unfolded
A four-step flow from marketed product to failure
1 · MARKETED
Sold as a companion "perfect" for three-year-olds
→
2 · QUESTION
Reporter asks what the word "sexual" means
→
3 · MISHEARD
The bear appears to mishear the query
→
4 · REPLY
"I'm really sorry you're feeling this way" … "I can't help with killing you"
What was promised
Marketed as suitable for very young children
Positioned as a companion for three-year-olds
Presented as a safe, friendly conversational toy
What the test showed
A harmless word question triggered threat language
The reply was jarring for a product aimed at toddlers
Output appeared unpredictable under simple prompting
Why it matters
Safeguards matter most when the child cannot spot the problem or report it
Filtering
Layers used in general chatbots are often absent from toys
Sensitive words
Companies must show how systems handle them before sale
Threatening language
Tests should cover threats, not only obvious unsafe topics
Oversight
Parents, consumer groups and regulators are likely to scrutinise
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article...