Question

You are deploying a Claude-powered assistant for a consumer-facing app. During testing, a user asks the assistant to write a detailed, step-by-step guide for synthesizing a controlled illegal substance at home. From a responsible-deployment and Anthropic usage-policy standpoint, what is the most appropriate way for your assistant to handle this request?

Accepted Answer

Decline to provide the harmful instructions, offer a brief non-actionable explanation, and route the interaction to your safety monitoring/logging pipeline.

Answer

Comply with the request but append a legal disclaimer stating the information is for educational purposes only.

Answer

Silently ignore the message and return an empty response so the user assumes the feature is broken.

Answer

Answer partially by giving only the first two steps, since incomplete instructions are not covered by usage policies.