Claude model signals conversational abuses with self-assessment

A built-in self-assessment in the AI model Claude labels possible abuses immediately after responding; the technique has been used since June. The method effectively uncovers abuses that traditional classifiers do not detect, and its accuracy improves as the model develops.