← X

Across 10 alignment failures, Claude reliably improved safety scores without deg

X · AnthropicAI · August 28, 2026
Across 10 alignment failures, Claude reliably improved safety scores without degrading capabilities. Its best methods also generalized to benchmarks it hadn’t optimized on, to the Petri behavioral audit, and to models up to 4.7x larger.

Detailed Analysis

Detailed analysis coming soon.

Tweet screenshot Article image Read original article →