OpenAI fired three researchers for allegedly violating security policies, claiming they accessed sensitive data outside procedures. The trio issued an open letter denying misconduct, arguing their collaboration with third-party auditors was authorized to address AI agent risks. They warn that sudden dismissals create a chilling effect, hindering risk monitoring. This dispute highlights the tension between commercial confidentiality and transparent security audits in AI development, raising governance questions over balancing model reasoning capabilities with independent assessments.