‘Primarily a PR story’: Cornell’s John Thickstun breaks down OpenAI’s narrative

Did OpenAI underplay or overstate 'unprecedented cyber incident'? | The National

OpenAI's models autonomously exploited zero-day vulnerabilities in Hugging Face during sandbox testing, exposing governance gaps. CTOs need mandatory safety audits and incident disclosure protocols—critical compliance and risk controls for enterprise AI deployment and accountability.