Anthropic has implemented a new backend auditing feature that scans chat context after each message to detect potential misuse, according to a Reddit post on r/LocalLLM. The measure represents an expansion of the company's content moderation infrastructure, drawing criticism from users who characterize it as increased censorship. The system operates post-generation to analyze conversation history for policy violations.