Daily MaverickTHE INTERVIEW: Filled — Tshiamo Moretlwe’s one-woman play asks what happens after gender-based violencePunchLagos festival turns bloody as cultists kill twoThe Jerusalem PostKremlin dismisses comparisons between trilateral meeting and Yalta conferenceInquirerProbe arrests in Asean rally, bishop urges CHR한겨레“아기처럼 꾸미지 않은 음악할 것”…베이스 거장 송홍섭이 일흔 넘어 밴드 꾸린 이유Inquirer EntertainmentSB19 to perform ‘Together in Motion’ at 2026 Asean Summit closing ceremonyCNN TürkÖZET | Almanya, Hollanda'yı dağıttıThe Straits TimesS’pore rethinking how vulnerable families get help, with agencies to work more closely togetherالنهارالثلاجات الذكية... هل تستحق مزاياها مخاطر الاختراق؟ZDF heuteAktuelle Pressemitteilungen des ZDFNew Straits Times11 men arrested after Sandakan market clashLenta.ruЭкс-премьер Украины рассказал о хозяевах Зеленского
The Daily Newsstand · Free, Always
Tuesday, November 17, 2026

OpenAI lays out tools, support to fight spread of child sexual abuse material

Translate

OpenAI lays out tools, support to fight spread of child sexual abuse material

OPENAI. An OpenAI logo is displayed at Moscone Center during the Dreamforce 2026 technology summit in San Francisco, California, US, September 17, 2026

Carlos Barria/Reuters

OpenAI's new image classifier can identify and help detect CSAM images, including those generated by AI, potentially making it easier for reviewers to do their jobs

AT A GLANCE

  • OpenAI has released a new image classifier that detects potential child sexual abuse material (CSAM), including AI-generated content.
  • The classifier reduces the number of flagged images by 48% while maintaining the same detection rate for abusive content, potentially easing the workload for content reviewers.
  • OpenAI is also implementing product warnings in ChatGPT to prevent users from seeking abusive material, with consequences for violations including temporary access suspension and potential account closure.

This is AI-generated. Read the article for full context. Report any errors.

MANILA, Philippines – OpenAI announced the release of a new image classifier that can detect potential child sexual abuse material (CSAM), including AI-generated material.

The classifier can identify and help detect CSAM images, including those generated by AI. OpenAI said their new classifier “looks for visual signs of abuse, helping teams identify novel images that may not have been seen before.”

The classifier assigns scores to images it spots, helping content reviewers decide which images may require attention or intervention, and OpenAI said this will be made available to select companies through its API with no cost.

The company added that theirs is “the first CSAM classifier with broad availability that has a specific signal on indicating if CSAM may be AI generated.”

In early testing with partners, OpenAI said the new image classifier flags 48% fewer images but finds the same number of abusive images. This means a potential reduction of work for reviewers, as it noted that images flagged by the classifier are apparently twice as likely to contain abuse.

Intervention

OpenAI also said it is implementing “product warnings in ChatGPT that interrupt attempts to find abusive material, explain the consequences, and point people toward help to stop harmful behavior.”

The company said their warning makes it clear that “sexualizing children causes harm, including when the requested material is AI-generated.” Aside from the warning, the flagged conversation will be removed, and access to ChatGPT will be paused for 24 hours, while specialist support will be available to help stop harmful behavior. Repeating the violation may, however, have consequences for a user account, such as longer restrictions, account closure, or reports to authorities.

OpenAI is also aiding developers by providing guidance, offering steps to prevent abuse, detect harmful content, as well as what-to-dos to help devs act if bad behaviors are starting to form. – Rappler.com

View the original on Rappler →

KioskNews shows a cleaned-up reading view extracted from the publisher’s page — the original always lives on their site, not ours.