OpenAI announced parental controls for ChatGPT on September 29 2025, but after testing them myself, I found critical flaws that make them ineffective for protecting children. In this video, I break down exactly what's wrong with the current system and what the parental controls should actually look like.
Key Problems Covered:
- Children can bypass controls by simply logging out
- Defaults are not safe-by-default (backwards safety engineering)
- Zero visibility for parents (not even conversation summaries)
- Controls are not tamper-proof
- System appears designed for PR, not actual child protection
What Real Controls Would Include:
- Tamper-proof architecture
- Safe-by-default settings
- Tiered monitoring (metadata + summaries, not full surveillance)
- Age-appropriate developmental stages
Do not rely on these controls. Kids should use ChatGPT only in shared spaces with ongoing parental engagement and clear family agreements.
