OpenAI releases CSAM image classifier model card
The story so far
OpenAI has released a CSAM image classifier model card trained and evaluated using controlled, manually labeled image datasets, which comply with US laws to prevent child sexual exploitation. The accuracy of online evaluations is significantly lower than offline evaluations because the proportion of CSAM in the online evaluation set is much lower than that of the offline set, leading to a small number of false positives possibly exceeding the number of actual detections, thereby reducing accuracy. OpenAI is subject to regulation by law, privacy, security, and reviewer well-being controls, and sensitive materials are stored and processed in a controlled environment with restricted access. OpenAI states that online evaluation results better reflect the labeled mix that human reviewers see on ChatGPT. OpenAI is providing the classifier to selected companies for free and is collaborating with the National Center for Missing and Exploited Children (NCMEC), Tech Coalition, and ROOST to share access. In early tests, the classifier reduced image labeling by 48% while marking the same number of abusive images, thereby reducing reviewer workload.
AI-written from the reports · Updated 5 hours ago
Developments
2 developments
- 17 Nov, 01:00 · 3 reportsA new image classifier is being releasedOpenAI News:OpenAI releases new tools and partnership support to strengthen the fight against online child sexual exploitation and abuse
- 16 Nov, 08:00 · 1 reportsOpenAI CSAM Image Classifier Model CardOpenAI Deployment Safety:OpenAI CSAM image classifier model card
Coverage timeline
Follow the coverage to see every side of the story.
- The Verge · AIOpenAI releases child sexual abuse image classifier to combat the proliferation of AI-generated content
OpenAI releases child sexual abuse image classifier to combat the proliferation of AI-generated content, which can identify abuse images including AI-generated content and reduce the false positive rate by 48%. OpenAI also provides free access to reviewed organizations and helps train law enforcement officers to use AI-assisted investigations.
- TechCrunch · AIOpenAI releases new image classifier and ChatGPT warning to combat child sexual abuse material
OpenAI announced today the launch of new tools to address growing AI tool safety concerns, including a new image classifier to detect child sexual abuse material (CSAM), warnings and temporary blocking measures to prevent users from creating or accessing sexualized child content, new guidelines for developers, and tools for law enforcement.
- OpenAI NewsOpenAI releases new tools and partnership support to strengthen the fight against online child sexual exploitation and abuse
OpenAI releases new tools and partnership support to strengthen the fight against online child sexual exploitation and abuse.
- OpenAI Deployment SafetyOpenAI CSAM image classifier model card
OpenAI has released the CSAM image classifier model card. The model is trained and evaluated using controlled, manually labeled image datasets that comply with U.S. laws to prevent child sexual exploitation. The accuracy of online evaluations is significantly lower than offline evaluations because the proportion of CSAM in the online evaluation set is much lower than that of the offline set, leading to a few false positives possibly exceeding the number of actual detections, thereby reducing accuracy.
Heat over time
- Now
- 29
- Peak
- 3417 Nov, 04:00
- 24h change
- –
Hover or tap the chart to see each hour; use the arrow keys on a keyboard.