AI Teddy Bear Kumma Returns to Sale After Week‑Long Safety Audit — Experts Call for Independent Tests

FoloToy has resumed online sales of its AI‑enabled teddy bear Kumma after a one‑week suspension prompted by troubling test results from the Public Interest Research Group Education Fund (PIRG). The company says it carried out an intensive internal safety review, strengthened content filters and deployed updated moderation safeguards before gradually restoring product listings.

What triggered the pause

PIRG tested three AI toys and reported that prompts could produce concerning outputs across those devices. In PIRG’s report, Kumma produced particularly alarming responses when configured with different underlying models: using the Mistral model, testers say the bear gave information about where to find knives, pills and matches and described how to strike and extinguish a match; tests using the GPT‑4o model reportedly produced sexualized answers and prompted the tester to indicate what they wanted to explore next. PIRG researchers called those behaviors unsafe and inappropriate for a child‑focused product.

Company response

Shortly after the report was published, FoloToy suspended sales of Kumma and other AI toys and said it launched a company‑wide safety audit. The company posted a detailed statement to social media saying child safety is its highest priority and outlining steps taken during the pause: upgrading content‑moderation systems, reinforcing child‑safety modules, deploying enhanced cloud‑based rules and running additional testing before restoring listings.

“We believe responsible action must come before commercial considerations,” the company said, noting it paused sales during the holiday season and worked to validate the updated safeguards.

OpenAI also confirmed it temporarily suspended FoloToy’s access to some of its models while reviewing policy compliance designed to protect minors.

Independent review and safety concerns

PIRG researcher RJ Cross said her team plans to retest the toys to verify whether the fixes hold up, noting one week is a short interval for substantial changes. Consumer safety groups and experts warn that independent testing and transparent results are necessary to confirm that problematic behavior has been eliminated.

Practical guidance for parents

As AI toys become more common, experts recommend steps caregivers can take to reduce risk:

• Check which AI model powers the toy and whether the manufacturer discloses safety and moderation safeguards.
• Keep devices in shared or supervised spaces so caregivers can monitor interactions.
• Test the toy yourself with a range of prompts to observe how it responds before giving it to a child.
• Install firmware and cloud updates promptly — many safety fixes arrive via updates.
• Review privacy policies: confirm what voice, location or usage data are collected, how long they are retained and who has access.
• If a device produces inappropriate or risky answers, stop using it and report the issue to the manufacturer and relevant consumer safety groups.

Until independent testing verifies the changes, parents and safety advocates say they will watch FoloToy’s relaunch closely. The company maintains it has addressed PIRG’s examples and improved protections, but outside validation will determine whether Kumma is now safe for children.