AI Round-up: 10 September 2026

AI safety dominated the window as AP reported researcher Jacob Coxon’s resignation from Anthropic with warnings about the pace of development. Separately, Anthropic published an assessment of unauthorised access during cybersecurity evaluations and announced an independent METR investigation. OpenAI called for mandatory national safety rules and appointed alignment researcher Paul Christiano to its Foundation Board and Safety and Security Committee.

Anthropic also released an interactive model of possible economic outcomes by 2030, exploring jobs, wages and growth through conditional scenarios. Video discussion examined OpenAI’s proposed Navier-Stokes proof and the dispute over research credit, without settling those claims. Elsewhere, Harvey reached a reported $15.5 billion valuation, Suno announced music models with Warner Music and BMG, NVIDIA expanded Australian AI infrastructure partnerships, and community posts compared coding assistants and questioned Perplexity’s paid-model access.