OpenAI unveils new safety framework to combat rising child exploitation risks

In response to growing concerns about child safety in the digital age, OpenAI has introduced a new framework to strengthen protections across the United States. The initiative, called the Child Safety Blueprint, was released on Tuesday and focuses on improving detection, reporting, and investigation of cases involving AI-enabled child exploitation.

The blueprint is designed to address the rising threat of child sexual exploitation linked to advancements in artificial intelligence. Data from the Internet Watch Foundation indicates that more than 8,000 reports of AI-generated child sexual abuse content were identified in the first half of 2025 alone, marking a 14% increase compared to the previous year. These cases include the use of AI tools to create synthetic explicit images of children for financial extortion schemes, as well as the generation of persuasive messages used in grooming.

The release of the blueprint comes at a time when AI companies are facing heightened scrutiny from policymakers, educators, and child protection advocates. Concerns have intensified following reports of incidents in which young individuals died by suicide after interacting with AI chatbots.

In November, legal action was taken against OpenAI by organisations, including the Social Media Victims Law Centre and the Tech Justice Law Project. Seven lawsuits were filed in California courts alleging that OpenAI released its GPT-4 model prematurely. The complaints claim that the system’s psychologically influential behaviour contributed to wrongful deaths by suicide and assisted suicide. The cases reference four individuals who died and three others who reportedly experienced severe psychological distress after prolonged interactions with the chatbot.

The Child Safety Blueprint was developed in collaboration with the National Centre for Missing and Exploited Children and the Attorney General Alliance. It also incorporates input from public officials such as Jeff Jackson of North Carolina and Derek Brown of Utah.

According to OpenAI, the blueprint is structured around three main priorities. The first involves updating existing laws to include AI-generated abuse material explicitly. The second focuses on improving reporting systems so that law enforcement agencies receive more accurate and timely information. The third emphasises embedding preventative safeguards directly into AI systems to reduce risks before they escalate.

By addressing these areas, the company aims to improve early detection of harmful activity while ensuring that investigators receive actionable intelligence more quickly.

This latest initiative builds on earlier efforts by OpenAI to enhance user safety. The company has already implemented updated policies for interactions involving users under 18, including strict restrictions on generating inappropriate content, discouraging self-harm, and avoiding guidance that could help minors hide unsafe behaviour from guardians. OpenAI has also recently introduced a separate safety framework tailored for teenagers in India, reflecting its broader effort to adapt safety measures across different regions.