Child Safety Online: New Rules for Vetting AI Training Data to Remove Illegal Content.
This bill aims to enhance child safety online by requiring AI developers to clean illegal content from the massive datasets used to train their systems. It mandates the creation of a voluntary framework by NIST, providing guidelines for detecting, removing, and reporting child pornography found in automatically collected training data. Citizens benefit from the assurance that major AI systems are built upon safer, vetted datasets, reducing the spread of harmful material.
Key points
NIST must develop a voluntary framework for AI developers to detect, remove, and report child pornography in training datasets.
Companies (AI developers and data collectors) following this framework receive limited legal protection from lawsuits related to these actions.
The National Science Foundation will support research into better methods for cleaning illegal content from AI training data.
Introduced
Citizen Poll