Anthropic proposes safety levels to manage dangerous AI capabilities

Original: Uk Ai Safety Summit

Source: anthropic.com ↗