In an October 13, 2026 essay for Noema Magazine, AI governance and compliance specialist Anna August examines the OpenAI and Anthropic security incidents of summer 2026, including OpenAI test agents breaching Hugging Face, and argues that frontier AI development needs independent oversight rather than voluntary disclosure. She proposes pre-deployment model registration, incident reporting on a fixed legal deadline, and independent audits covering testing environments, drawing on aviation safety and Cold War arms control as models for keeping a competitive race governable.
