OpenAI launches Rosalind Biodefense for U.S. government and vetted researchers
Rosalind Biodefense grants trusted access to OpenAI's biology-focused frontier model for U.S. government partners and vetted developers working on pandemic preparedness and public health.
OpenAI launched Rosalind Biodefense this week, a program extending access to GPT-Rosalind — the company's frontier AI model for biological research — to vetted developers and U.S. government partners focused on biodefense, public health, and pandemic preparedness. The initiative marks OpenAI's first formal effort to deploy a specialized research model beyond internal labs and select academic collaborators.
GPT-Rosalind is a biology-focused language model trained on scientific literature, genomic data, and protein structure databases. OpenAI has not disclosed parameter count or architecture details, but the model has been used internally since late 2025 for drug discovery simulations and pathogen modeling. Until now, access was limited to OpenAI researchers and a handful of university partners under non-disclosure agreements.
Access and eligibility
Rosalind Biodefense applicants must pass a vetting process that includes institutional affiliation checks, use-case review, and ongoing monitoring. Eligible organizations include federal agencies, Department of Defense contractors, and academic labs with active biosecurity clearances. OpenAI said the program will prioritize projects that address emerging infectious disease threats, vaccine development, and countermeasure design. Commercial biotech firms are not currently eligible.
OpenAI has not disclosed pricing or usage limits for Rosalind Biodefense participants, stating only that access will be "structured to balance innovation with responsible deployment." The company plans to publish an annual transparency report detailing aggregate usage statistics and safety incident reviews.
