The message from OpenAI states that this early access program complements the existing process for testing advanced models. It includes thorough internal safety testing, external teams such as the Red Teaming Network, and collaboration with various private and public organizations.
OpenAI hopes that this testing will provide new insights, deepen the understanding of risks, develop new assessments, and highlight areas for advancing research in safety.