AI Tutor Safety: How Koydo's Everyday Tutor Is Designed Around Child Protection
The product-design principles Koydo uses to keep its AI tutoring features safe for children — including the rules around voice processing, prompt isolation, and adult-only categories.
Synthesized from Koydo's internal AI Safety Policy, the public Privacy Policy, and the COPPA Notice. Cross-referenced with Anthropic, OpenAI, and Google AI provider safety documentation.
No advertising profile
Koydo does not run advertising and does not build advertising profiles from tutor sessions. AI-tutor conversations are processed for the immediate session and used only to support the learner's stated educational task.
Process-and-discard voice (under 13)
Pronunciation practice for under-13 learners sends voice clips to the AI provider in real time and discards them immediately. Koydo does not retain a biometric template, voice fingerprint, or audio recording from these sessions.
Age-aware prompts
The Everyday Tutor receives a different system prompt for each age bracket. Younger learners get a stricter content posture, simpler vocabulary, and a refusal to engage with adult topics. Older learners get more depth, including the ability to discuss sensitive academic topics where pedagogically appropriate.
Human review path
Any AI-suggested action that would affect billing, account access, eligibility, or legal rights is escalated to a human reviewer. The AI tutor itself never modifies these settings. The applicable user-facing policy is published at /legal/automated-decisions.
Provider diversity
Koydo's tutor stack uses multiple AI providers (Anthropic, OpenAI, and Google) in rotation for resilience and to avoid single-vendor lock-in. Each provider operates under a contractual data-processing agreement and Koydo's child-safety addendum.