How Friday’s safeguards work
The safeguards currently implemented—and their limits.
Shared instructions
Text and voice use shared age-aware safety rules; pictures receive age-aware educational instructions. This is guidance, not a guarantee of safe output.
Request checks
Some unsafe text patterns are blocked before a lesson starts. Disclosures of danger receive support-first handling. Pattern checks can miss requests or misunderstand context.
Permission checks
New AI requests require a signed-in parent and learning permission. Voice also needs separate permission and a short-lived one-use connection ticket.
Reporting and review
Families can save reports and view receipts. Educator review requires an assigned reviewer role; approvals cover samples, not every AI response.
What is still needed
Qualified safeguarding review, ongoing evaluations, provider child-processing approval, final retention practices and live voice testing remain needed before commercial launch.