Safety & Design
AI safety
The work of reducing harm from an AI system: dangerous replies, sexual content, distress that never reaches an adult, and tools a child should not have.
- Parents
- Educators and school leaders
What parents should know
AI safety is how a product limits harm before and after a bad prompt. Filters help. They miss things. A parent who can read the chat is the backup. A safety page that promises a perfect block is selling a wish.
On this page
What is AI safety?
AI safety covers what the system will not do and what happens when something goes wrong. Topic limits, age filters on images, refusals of dangerous how-tos, and a path for distress are safety controls. NIST's risk framework is a way to name those controls. It is not a seal on a product.
For children, safety also includes the account. A hidden chat that a parent cannot open is a safety gap even if the model refuses some words. The transcript is part of the control, not an extra.
Why AI safety matters
Kids rephrase a blocked question. They also paste scary feelings into a box that always answers. A filter without an adult, or an adult without a filter, is incomplete. You want both, and you want the product to say what it will not automate.
Image tools need their own line. Age filters, and a refusal to fake a real person's body or voice, are safety decisions. They are separate from whether the chatbot can help with a math step.
How it shows up in practice
- A dangerous how-to is refused, and the parent can still see the attempt.
- A distress message shows a crisis resource and alerts the parent.
- Image generation can be turned off for one child.
How HeyOtto helps
AI safety on HeyOtto is per-child topic limits, age-filtered image tools, and a parent who can read the chats. If a child shows distress, the chat shows crisis resources, including 988, and the parent gets a real-time alert. HeyOtto does not auto-contact emergency services, and it does not claim a filter never fails.
- HeyOtto does not offer face swap, a voice clone of a real person, or nudify tools.
- A parent can tighten topic limits or turn image generation off.
- There are no ads.
For families
Try freeFAQs
Is a content filter the same as AI safety?
A filter is one control. Safety also includes who can read the chat, what happens in distress, and which tools exist at all. HeyOtto uses limits plus a parent transcript.
Will the app call 911?
No. HeyOtto does not auto-contact emergency services. You get an alert, and the chat can show a crisis resource such as 988. If someone is in immediate danger, you make that call.
Can kids get around the limits?
Sometimes. That is why the chat stays visible. HeyOtto does not publish bypass prompts, and it does not claim every attempt is caught.
Sources
Last reviewed September 26, 2026. This entry is reviewed twice a year.