OpenAI has introduced a classifier for identifying potentially child sexual abuse material, including AI-generated images, and has strengthened ChatGPT’s warnings and interventions in response to harmful requests. Reviewed organizations can use the tool free of charge through the API; early evaluations suggest it could reduce by 48% the number of images requiring human review, while detecting a similar number to existing tools. When ChatGPT identifies such a request, it removes the conversation and suspends access for 24 hours. Repeated violations may result in longer-term restrictions, account closure, or referral to the authorities.
Source:Digital Trends · Read original