OpenAI announces new filter against abusive material
The company links an image classifier to warnings, blocks and reports of suspected violations.
Follow-up to: Reports of AI-generated child sexual abuse material soar Monday, 16 November 2026, 05:07
Listen to this article
There is no audio version yet. Request one and an AI voice will read the article aloud.
Read by an AI voice.
OpenAI announced new safety measures on Monday against material involving the sexual exploitation of children. According to the company, the approach consists of a new image classifier and stricter warnings and blocks in ChatGPT.
The new classifier is intended to recognise images that fall into the prohibited category. OpenAI also says it will issue warnings and temporarily block accounts when users try to create or access prohibited material. In cases of serious or repeated violations, an account may be permanently closed.
According to reports on the announcement, the classifier was developed with help from the National Center for Missing and Exploited Children, a US organisation that processes reports of online exploitation. OpenAI also says it is working with the Tech Coalition and ROOST to make safety tools available to other parties.
The measure comes after reports of AI-generated abusive material rose sharply. That earlier figure says something about reports and not automatically about the total number of files created or distributed. Nor is every image flagged by an automated system illegal; human assessment and legal context remain necessary.
OpenAI says it reports suspected prohibited material to the relevant reporting authorities. The announcement and help information contain no independent error or success rates.
The development therefore shows two sides. Automated recognition can help platforms intervene more quickly, but filters can also wrongly flag legitimate material and raise questions about privacy, transparency and the transfer of data to law-enforcement or reporting authorities. The new tool has been announced, but independent figures on its effectiveness are not yet available.
Also show the region and a close-up
One story, several perspectives
What is established
- OpenAI has announced new detection and enforcement measures.
- The company says it uses warnings, temporary blocks and potentially permanent account closures.
- Independent figures on the error margin and effectiveness have not been published.
Left
Arguments A possible left-wing perspective emphasises that platforms have a major responsibility to protect children and that voluntary measures should be verifiable.
Values Child protection, public responsibility and the protection of vulnerable groups.
Consequences More oversight and transparency may be necessary, but too many automated blocks could affect legitimate users.
Centre
Arguments Supports targeted detection combined with human assessment, clear procedures and independent audits.
Values Safety, proportionality, legal protection and practicality.
Consequences A combination of technology and human oversight can limit the risks, but takes time and money.
Right
Arguments Emphasises that platforms should remove prohibited material quickly and that users who knowingly violate the rules should face firm action.
Values Enforcement, user responsibility and the protection of order and safety.
Consequences Stricter sanctions may act as a deterrent, but errors in automated systems call for caution over permanent measures.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The measures described can be found in OpenAI’s documentation and in independent reporting. Claims about effectiveness have deliberately not been presented as established facts.
- confirmed OpenAI announced new measures against abusive material on 16 November. — The announcement and publication date are stated in TechCrunch’s reporting. source
- confirmed The measures include an image classifier, warnings and blocks. — These components are listed in the reporting. source
- confirmed OpenAI can close accounts for violations. — This is stated in the official Help Center information. source
- confirmed Independent effectiveness figures are lacking. — The announcement and help information consulted publish no independent error or success rates. source
Editor's note
This follows the earlier report on rising numbers of reports. The new measures were announced by OpenAI; independent effectiveness figures are lacking.Sources
The story so far
- Monday, 16 November 2026, 05:07 Reports of AI-generated child sexual abuse material soar
- Monday, 16 November 2026, 20:47 OpenAI announces new filter against abusive material (this article)
More on this in Dutch media
- AD — „openai chatgpt”
- RTL Nieuws — „openai chatgpt”
- FD — „openai chatgpt”