Anthropic Introduces Constitutional AI 2.0 for Safer Model Training

Anthropic publishes a major update to its Constitutional AI training methodology, demonstrating improved safety alignment without sacrificing capability on downstream tasks.