Anthropic lays out vision for bolstering AI model safety

The policy framework emphasizes evaluation and deployment safeguards. Anthropic proposes industry standards to address catastrophic risks from advanced AI.