🆕@AnthropicAI's Claude Fable 5.2 checks the data before it trusts it. We ran it against Claude Fable 5.1 on real enterprise document tasks: pull the source files, reason across them, produce the actual deliverable.
→ Public sector: 92% vs 81% (+11 points)
→ Energy: 93% vs 85%





