Generative AI (GenAI) is transforming our world, offering incredible opportunities for innovation. However, this technology also presents new and complex challenges, especially concerning online child safety. A critical new blueprint, "Protecting Children in the Age of Generative AI," outlines a comprehensive framework to address the misuse of GenAI to facilitate child sexual abuse material (CSAM) and exploitation.
This blueprint represents a significant step forward, aligning the efforts of technology providers, law enforcement, and advocacy groups.
Foreword from Leadership
The blueprint is introduced by State Attorneys General Jeff Jackson (North Carolina) and Derek Brown (Utah), Co-Chairs of the Attorney General Alliance's AI Task Force. They emphasize the need for proactive, adaptive strategies:
"We are particularly encouraged by the framework's recognition that effective GenAI safeguards require layered defenses — not a single technical control, but a combination of detection, refusal mechanisms, human oversight, and continuous adaptation to emerging misuse patterns... Getting the prevention architecture right upstream is the single highest-leverage investment the industry can make in child safety."
Karen White, Executive Director of the Attorney General Alliance, and Michelle DeLaune, President & CEO of the National Center for Missing & Exploited Children (NCMEC), also applauded the initiative, stressing the importance of collaboration across all sectors to reduce harm and support children's safety.
The Evolving Threat
While digital services have unfortunately long been misused by bad actors, generative AI introduces specific new dynamics that strain existing legal and investigative systems. These threats include:
- Synthetic CSAM: AI can be used to create realistic, entirely synthetic depictions of abuse without a direct victim.
- Digital Alteration: Existing imagery can be easily manipulated.
- Scale and Speed: Offenders can operate more efficiently across different content formats (text, image, video).
The Policy Blueprint: Three Reinforcing Priorities
The framework advanced in this document focuses on three mutually reinforcing pillars designed to cover the full lifecycle of harm—from prevention and detection to investigation and prosecution.
Priority One: State Legislative Modernization
The goal is to ensure that state laws remain fully enforceable and effective as technology evolves. Key recommendations include:
- Updating CSAM Definitions: Explicitly covering AI-generated and digitally altered material.
- Clarifying Attempt Liability: Ensuring that intentional attempts to generate abusive material remain prosecutable, even if safeguards block the output.
- Establishing Good-Faith Safe Harbors: Protecting providers who conduct responsible detection, reporting, and safety research from unintended liability.
According to research cited in the blueprint, as of August 2025, 45 states have already enacted laws addressing AI-generated or computer-edited CSAM, underscoring widespread legislative concern.
Priority Two: Best Practices for Provider Reporting & Coordination
This section aims to improve the quality and actionability of reports made to NCMEC’s CyberTipline. Recommendations include:
- High-Quality, Structured Reports: Providing complete details (Who, What, Where, When) rather than just file excerpts.
- AI-Assisted Triage with Human Review: Using AI to surface high-risk activity but maintaining human oversight for reporting decisions.
- Reducing Investigative Burden: Bundling reports by user or incident and including technical identifiers (hashes, IP port numbers) to connect related activity quickly.
Priority Three: Safety-by-Design Prevention & Detection Safeguards
The most effective way to protect children is to interrupt exploitation attempts before harm occurs. The blueprint calls for:
- Intent Detection: Detecting high-risk prompts and behavioral patterns.
- Generation Refusal: Systems must actively refuse prohibited requests and implement intervention mechanisms (like friction or throttling).
- Continuous Risk Monitoring: Regularly evaluating and adapting safeguards to address emerging misuse patterns.
Conclusion
Protecting children online is a shared responsibility. The rise of generative AI demands updated legal frameworks, improved reporting mechanisms, and robust safety safeguards built directly into the technology. This blueprint provides the roadmap for government, law enforcement, non-profits, and the tech industry to collaborate effectively and ensure innovation supports child safety.
