
Anthropic Says Claude Reduced Alignment Failures In Automated Tests
Anthropic said Claude improved safety scores across 10 alignment failures and trained smaller models without reducing their general capabilities in tests.
Murtuza Merchant6 hours ago


































