OpenAI said Monday it is adding a warning in ChatGPT when someone tries to find or create material that sexually exploits children. The company says it will remove the flagged conversation and pause access to ChatGPT for 24 hours, while keeping access to specialist support. It also released a separate tool that helps safety teams identify potentially abusive images.

The announcement describes two changes with different audiences. ChatGPT users may encounter the warning inside the product. Vetted organizations can apply to use the image detection tool at no cost through OpenAI’s developer platform. For families trying to report exploitation or stop an image from spreading, the National Center for Missing & Exploited Children already offers services they can reach directly.

What changes inside ChatGPT

OpenAI already prohibits child sexual abuse material, trains ChatGPT to refuse requests for it and can take action against accounts that break its rules. The new warning adds an intervention that explains the harm and directs users to help. OpenAI says that includes attempts involving AI-generated material. The announcement gives no country-by-country rollout schedule or date when all users can expect the change.

Who can use the image detection tool

Glass office building at 1515 Third Street in San Francisco.
The San Francisco building that housed OpenAI’s headquarters when this photo was taken in June 2025.Coolcaesar · CC BY 4.0

OpenAI developed the tool with assistance from the National Center for Missing & Exploited Children. It flags images for review, including potentially abusive material generated by AI. Safety teams can use it to help identify images their systems have never seen before. OpenAI says organizations need to request access and go through vetting.

In an evaluation using images uploaded by users, OpenAI estimated that the new tool would flag 48% fewer images while finding roughly as many abusive ones as its existing tool. The estimate describes the volume of flags in that comparison. OpenAI did not publish a figure for how much abuse the tool will detect across other services. Its developer guidance describes how companies can review flags and handle reports.

Where families can report abuse

Anyone in the U.S. can report suspected online child sexual exploitation to NCMEC’s CyberTipline. The center tells people to call 911 or local police if someone is in immediate danger.

NCMEC’s free Take It Down service can help limit the spread of an explicit image or video taken when the person depicted was under 18. The service creates a digital fingerprint on the person’s device and shares the fingerprint with participating platforms, so they can look for copies. The image or video stays on the device. NCMEC tells users to use a file they already have and to avoid downloading, sending or sharing one to submit it.