Top AI models fail commercial underwriting with under 3% accuracy

Original: Frontier models score <3% on commercial underwriting benchmarks.

Source: x.com ↗