AI and children’s safety

Review work and supervise children.

AI limitations

Frikkie can produce plausible but incorrect answers, unsafe code, faulty calculations, misleading sources and designs that resemble existing work. Review and test outputs before use. Research availability and source quality vary. No promise is made that the service will prevent every harmful response or eliminate legal risk.

Children’s learning pilot

The child area provides reviewed learning activities. Parents can separately enable supervised AI questions with input and output checks. Questions and recent in-page context go to Groq for generation and checks. The children’s service allows five question attempts per child and twenty across all children per UTC day. AI filters can fail; parents should supervise. No web search, image generation, uploads or purchases are available through child codes. Child codes cannot authorise adult APIs. Parents should keep adult account credentials private, sign out on the child’s device, supervise use and revoke access when required. Parental authority is self-declared and is not independently verified.

Help and reporting

If a child is worried, encourage them to speak with a parent or trusted adult. Report service problems to the support contact. Use appropriate local emergency services for immediate danger.

Future changes

Expanded child AI features require further safety testing and updated parental information before release. A more advanced Frikkie is in development. Current features, capacity and policies can change, with updated notices and acceptance where required.

Operator and contacts

Vra_Frikkie 40 3rd Street, Northmead, Benoni, South Africa support@vrafrikkie.co.za support@vrafrikkie.co.za