← Voltar ao feed
low

OpenAI launches CSAM image classifier, ChatGPT deterrent warnings with 24-hour suspensions and law enforcement tools

1 fonte(s) · 1 atualização(ões) · publicado 16/11/2026 14:00 · coletado há 1 dia

OpenAI announced a set of child-safety measures against child sexual abuse material (CSAM): a new image classifier that detects CSAM, including AI-generated imagery, plus warnings and temporary blocks in ChatGPT, developer guidelines and tools for law enforcement. The classifier was developed with NCMEC. Access is shared through Tech Coalition and ROOST, and select companies get it free via the API. This is product and policy news with no direct threat to defenders. The research step could not be performed, so the report relies on the article alone.

Detalhes técnicos

  • The classifier scores previously unseen images to guide reviewers. In early partner testing it flagged 48% fewer images while finding the same number of abusive images, and flagged images were twice as likely to be abusive (techcrunch.com).
  • ChatGPT now shows a warning and points users to support when they attempt prompts that sexualize children. It suspends the account for 24 hours, and repeat offences can lead to permanent removal (techcrunch.com).
  • The deterrent messaging was shaped with Protect Children, the Lucy Faithfull Foundation and Joel Scanlan of the University of Tasmania (techcrunch.com).
  • OpenAI also published guidelines to help developers prevent, detect and report CSAM, and is giving law enforcement AI plugins, investigative tools, training and a field guide (techcrunch.com).

Linha do tempo

  1. techcrunch.com ↗ · pub. 16/11/2026 14:00 · coletado 16/11 15:00 primeiro registro

    OpenAI launches CSAM image classifier, ChatGPT deterrent warnings with 24-hour suspensions and law enforcement tools

    ler a análise desta fonte

    OpenAI announced a set of child-safety measures against child sexual abuse material (CSAM): a new image classifier that detects CSAM, including AI-generated imagery, plus warnings and temporary blocks in ChatGPT, developer guidelines and tools for law enforcement. The classifier was developed with NCMEC. Access is shared through Tech Coalition and ROOST, and select companies get it free via the API. This is product and policy news with no direct threat to defenders. The research step could not be performed, so the report relies on the article alone.

    Technical Details

    • The classifier scores previously unseen images to guide reviewers. In early partner testing it flagged 48% fewer images while finding the same number of abusive images, and flagged images were twice as likely to be abusive (techcrunch.com).
    • ChatGPT now shows a warning and points users to support when they attempt prompts that sexualize children. It suspends the account for 24 hours, and repeat offences can lead to permanent removal (techcrunch.com).
    • The deterrent messaging was shaped with Protect Children, the Lucy Faithfull Foundation and Joel Scanlan of the University of Tasmania (techcrunch.com).
    • OpenAI also published guidelines to help developers prevent, detect and report CSAM, and is giving law enforcement AI plugins, investigative tools, training and a field guide (techcrunch.com).