← Back to the blog

OpenAI's New Tools Against AI-Generated Child Abuse Images: What Parents Need to Know

By Equipo IA University Kids · Published · Updated

Leer este artículo en español

What's new

On November 16, 2026, OpenAI published Strengthening the fight against online child sexual abuse with new tools and support. Here are the four pieces:

  1. A CSAM image classifier. It looks for visual signs of abuse, including in brand-new images and AI-generated material. OpenAI says it's the first broadly available CSAM classifier with a specific signal for whether content may be AI-generated. It was trained and evaluated with help from the National Center for Missing & Exploited Children (NCMEC) and is offered free to vetted organizations through the OpenAI API. Technical details are in the model card.
  2. Warnings in ChatGPT. When someone tries to find material that sexualizes children, ChatGPT shows a warning explaining the harm (including when content is AI-generated), removes the conversation, pauses ChatGPT access for 24 hours, and points to specialist help to stop the behavior. Repeat violations can lead to longer restrictions, account closure or reports to authorities.
  3. Guidance for developers, built with input from Thorn, NCMEC, the Tech Coalition and the Internet Watch Foundation, to help other apps prevent, detect and report abuse.
  4. Tools and training for child-safety law enforcement units.

How well does it work?

These are OpenAI's own and early partners' results, not published independent evaluations:

  • Compared with OpenAI's existing protections, the new classifier would flag 48% fewer images while finding roughly the same number of abusive ones, and its flags were about twice as likely to contain abuse.
  • In testing by fal, reviewers confirmed 72% of high-confidence flags as novel CSAM; in SafetyKit's evaluations of drawings or animations that sexualize children, it caught 90% of cases.

Why it matters for parents and teachers

AI "nudify" tools and fake sexual images made from real photos of kids and teens are a real risk. A major AI company sharing a free detection tool with other platforms could help that content get found and taken down faster.

But keep the limits in mind:

  • The classifier is for platforms and organizations, not families. It's not a parental control app.
  • ChatGPT's protections don't cover other photo-editing or AI apps your child or classmates might use.
  • No tool replaces a child who feels safe telling you what happened.

If it happens: what to do and where to report

  • Don't forward, download or share the image, not even "as evidence" or to show someone. That spreads it further.
  • Report it on the platform where it appeared (social app, chat, game).
  • In the U.S.: Take It Down, a free NCMEC service, helps stop the spread of nude or sexually explicit images taken of someone before they turned 18. It creates a digital fingerprint (hash) without the image ever leaving your device, and participating platforms use it to find and remove copies. You can also report to NCMEC's CyberTipline.
  • In Colombia and Mexico: Te Protejo takes confidential reports involving minors, with a focus on sexual violence in digital spaces.
  • If a child is in immediate danger, call 911 or your local emergency number.
  • Above all, support your child: it is never the victim's fault.

Try this at home: "My image, my choice"

A 15-minute, age-appropriate conversation (roughly 10 and up):

  1. Explain calmly: AI can now alter photos to make it look like someone did or showed something that never happened.
  2. Ask: "Have you seen or heard about anything like this at school or in your group chats?" Listen without punishing.
  3. Agree on three rules: don't post or pass along photos of others without permission; never forward an intimate or fake image of anyone; and if you receive one or someone threatens you, tell a trusted adult.
  4. Make the most important promise: "If something happens, you won't get in trouble for telling me. We'll handle it together."

For teachers: add AI-manipulated images to your digital citizenship lessons: what they are, why sharing them causes real harm, and who students can go to at school.

Sources

  1. Strengthening the fight against online child sexual abuse with new tools and support (OpenAI) · 2026-11-16
  2. OpenAI CSAM Image Classifier - Model Card (OpenAI Deployment Safety Hub) · 2026-11-16
  3. Take It Down (National Center for Missing & Exploited Children, NCMEC) · consultado 2026-11-18
  4. CyberTipline - Report Child Sexual Exploitation (NCMEC) · consultado 2026-11-18
  5. Te Protejo Latino - reporte confidencial (Colombia y México) · consultado 2026-11-18

Want your kid to learn AI with guidance and purpose?

Discover the AI course for kids and teens ages 8 to 16.

See the course

Comments

No published comments yet.

Please don't share personal information about minors (full names, school, exact age, photos or location). Comments are reviewed before they're published.