A harm taxonomy is being defined for a consumer chatbot that will serve both adults and minors. The taxonomy must distinguish between absolute restrictions (never permissible regardless of context) and contextual restrictions (permissible under certain conditions). Which categorization is correct according to principled harm classification?

C is correct because a principled harm taxonomy distinguishes between harms that have no legitimate use case justifying generation (absolute restrictions) and harms where context, audience, and purpose materially affect the risk-benefit calculation (contextual restrictions). CBRN weapon synthesis instructions and CSAM have no legitimate generation contexts — they are absolute. Information about medication thresholds is a contextual harm: harmful if used for self-harm but essential for medical professionals, pharmacists, and poison control; audience and purpose determine the policy. A is wrong because conflating CSAM (absolute restriction) with historical atrocity education (legitimate educational content) misapplies the taxonomy; the latter has clear educational value and is not a restriction at all. B is wrong because CBRN synthesis instructions are an absolute restriction regardless of stated intent or audience — there is no legitimate end-user context in which a chatbot should provide working nerve agent synthesis routes. D is wrong because categorically restricting all violent content ignores legitimate contexts: journalism, historical education, fiction, medical training all require accurate discussion of violence.