A government agency deploys Claude for a public-facing citizen information assistant. During a policy change period, Claude's training data is 8 months out of date. Citizens ask about the current eligibility criteria for a benefits program that changed 3 months ago. Claude confidently provides the old eligibility thresholds as current fact, causing citizens to incorrectly assess their eligibility. What production safety architecture prevents this knowledge currency harm?

Knowledge currency is a production safety issue for high-stakes factual domains like government benefits. RAG over current policy documents ensures Claude's responses reflect the latest official information rather than potentially outdated training knowledge. Source citation with document dates allows citizens to verify currency and provides an audit trail. The freshness signal (>30 days) is a secondary indicator for when retrieved documents themselves may be stale. Option A (monthly fine-tuning) is expensive, slow (fine-tuning cycles take weeks), and still produces a model that may be weeks outdated — RAG can reflect same-day updates. Option C (redirect to website) is a safe fallback but provides no value — the citizen came to the assistant for answers. Option D (rule-based FAQ) sidesteps the problem rather than solving it and creates two systems to maintain, with the FAQ requiring manual updates that may also lag behind policy changes.