Anthropic Risk Report: August 2026

AI lab Anthropic has released a detailed risk report covering events through mid-July twenty twenty-six, as reported by Zvi Mowshowitz. The document reveals the existence of an internal-only system called Model Two, which outperforms previous versions in substituting for human AI researchers. While the report provides transparency on risks involving autonomy and biological weapons, Mowshowitz notes that the lab's risk thresholds have become more complex and harder to parse.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton