Anthropic's Opus 5 System Card Quantifies Safety Gap With and Without System Prompts

Anthropic's Claude Opus 5 system card, published July 24, 2026, discloses that the model's harmless response rate on the API drops slightly below recent model baselines when no system prompt is provided. This is among the clearest vendor disclosures quantifying that safety properties depend on scaffolding, not just the model itself. API users overriding default prompts should factor this delta into risk assessments.