OpenAI is addressing recent third-party cybersecurity evaluation incidents involving its AI models and introducing new safeguards to strengthen model testing and evaluation processes. The announcement aims to build trust in the security of frontier AI systems as evaluation standards mature across the industry.