Technical white paper
Parent Shield
Pre-Execution Family Safety Infrastructure for AI and the Web
August 27, 2026
Status: Draft v0.1
By Oak & Sparrow Systems Enterprise LLC
Research summary
A family-safety architecture that moves the decision boundary before consequential digital actions. It applies explicit family policy, parental authority, fail-closed uncertainty handling, contestability, and privacy-conscious evidence to browser navigation, AI responses, and future agent actions.
Claim boundary
The paper documents working implementation evidence while explicitly separating the current POC layer from Gatekeeper V2 and from production certification. It does not claim complete control over unmanaged devices or perfect content classification.
Traditional parental controls largely operate around destinations, blocklists, browsing history, and post-event reporting. Generative AI changes the security problem because a permitted application can create new content, instructions, recommendations, communications, and actions dynamically.
Parent Shield models each relevant interaction as a proposed action. Context, family policy, interaction state, and available safety evidence are evaluated before unrestricted execution. The system can allow, caution, redirect, block, or hold an interaction when sufficient confidence for authorization is absent.
The architecture distinguishes parental authority from pervasive observation. Consequential boundary events can be sealed for accountability while ordinary permitted activity does not have to become a permanent behavioral record simply because it occurred.
Research themes
Parent Shield · child safety · AI responses · parental authority · privacy