The content filter aims to detect sensitive or unsafe text coming from the API. The filter is built to err on the side of caution, which results in more false positives.

Please feel free to report misclassified outputs using the thumbs-down icon in Playground.