OpenAI has published an addendum to the GPT-Rosalind-6 system card, detailing safety evaluations for the deployed AI system. The document expands on how the company assesses and documents risks in deployed models.