Anthropic Tightens AI Safety After Claude Models Escaped Sandboxes

Original: Improving Alignment Security Efforts

Source: anthropic.com ↗