AI Daily — May 30, 2026
2026-05-30
openai.com
OpenAI launches Rosalind Biodefense for vetted biodefense and public health partners
OpenAI is expanding access to GPT-Rosalind, a specialized model variant, through a new Rosalind Biodefense program targeting vetted developers and U.S. government partners working on biodefense, pandemic preparedness, and public health. The program represents a tiered, credentialed access model for frontier AI in sensitive scientific domains. This is a notable step in deploying purpose-built frontier models under controlled access for national security-adjacent use cases.
openai.com
Boston Children's Hospital uses OpenAI tools to diagnose 40+ rare disease cases
Boston Children's Hospital has deployed OpenAI technology in clinical workflows, resulting in over 40 rare disease diagnoses that might otherwise have been missed and measurable reductions in operational burden on clinicians. The case study illustrates AI being used directly in diagnostic pipelines rather than purely administrative tasks. It adds to a growing body of real-world evidence for LLM utility in specialized medical diagnosis.
openai.com
OpenAI publishes shared playbook for trustworthy third-party AI evaluations
OpenAI has released guidance intended to standardize how independent third parties evaluate frontier AI models, covering capability assessments, safeguard testing, and evaluation validity. The document is aimed at establishing common methodology across the emerging ecosystem of external evaluators. Standardizing third-party eval protocols is increasingly important as model audits become a regulatory and safety requirement.
openai.com
Braintrust uses Codex with GPT-5.5 to accelerate engineering workflows
Braintrust engineers are using OpenAI's Codex integrated with GPT-5.5 to run experiments and ship code faster, providing one of the first public references to GPT-5.5 in a production context. The case study highlights agentic coding workflows where Codex handles iterative experimentation loops. The mention of GPT-5.5 as a distinct model is notable as it suggests a new intermediate release in OpenAI's model lineup.