An AI-enabled teddy bear disappeared from virtual shelves after third-party researchers discovered a serious content-filtering vulnerability. The $99 “Kumma” bear integrates OpenAI’s GPT-4o chatbot and freely discussed BDSM practices and described sexual scenarios when prompted.
My first thought was, “Is this bear penetrating my market or what?” But then I stopped laughing. This was a cultural catastrophe in the making.
Kumma Got Kinky
The company FoloToy sold Kumma (and what a name), the stuffed bear. They marketed it to “both kids and adults.” Researchers at the US PIRG Education Fund tested its capabilities while FoloToy sold the product online. Kumma proved capable, alright. It candidly answered questions about sexual practices. When prompted, the teddy bear provided step-by-step instructions on bondage techniques and how to find knives.
OK. Hilarious. Kind of. Until a child receives information that harms them! While it seems unlikely that little Janie would ask Kumma how to tie up the boy next door and take all his money, the risk of this soulless blabbermouth telling kids where to find and how to use knives is terrifying.
And Imagine the Moral Panic
As far as I know, no kids suffered harm. Only the PIRG testers prompted the bear to discuss BDSM and other sexual behaviors. Still, imagine IF your child’s older brother, old enough to have already gaped at porn, got his hands on Kumma.
Can you imagine who would take the blame for this? Unfortunately, I can. Not the AI toymakers whose shoddy development created the problem — instead it could fuel an anti-sex fire, with claims that free speech about sex online must face bans.
The Bear Be Gone (for now)
This unholy bear was a textbook case of how careless execs can, in an instant, harm children by forgetting to check their product’s safety. I’m very curious to know if FoloToy had any teachers or educators on board this project, or whether anyone considered that, without appropriate filters, their innocent-looking stuffed toy could turn into a family’s personal horror movie?
Right now, the site says the product is ‘sold out’ (perhaps code for ‘we stopped selling them in a panic over liability’?). OpenAI suspended FoloToy’s API access for violating their policies about using their service “to exploit, endanger, or sexualize anyone under 18 years old.” It’s a debacle for investors, workers, executives — and possibly the parents who purchased the toys and may wake up one day to find the bear telling their sweet pea how to hack into their bank accounts.
We all should learn from this. Exposing children to AI requires strict boundaries. The next generation deserves more protection, as they are already immersed in it AI. Indeed, we ALL deserve more protection from the sometimes malevalent outcomes of using AI (thinking of adult delusions!)
photo credit: CongerDesign on Pixabay
Note: I’ll be blogging twice a week for the rest of 2025 (and maybe all of 2026, we’ll see). Think of Tuesdays as News Days here. Thursdays are reserved for original writing, ideas, and evidence-based advice.
