OpenAI builds bio-risk evals for frontier models, openly seeks critics
iskander · x · 2026-09-16
James Black, who joined OpenAI a few months ago, announced he is building the company's evaluations to understand biological risks from its models, and publicly invited biosecurity experts on both sides of the debate to weigh in where views are underrepresented.
Stanford's Anshul Kundaje amplified the call, noting the team is open-minded and genuinely working to build useful rule-out/rule-in evals and scenarios for frontier models.
Related event: OpenAI Builds Biorisk Evaluation Team, Seeks Expert Input(2 posts)→
More from Companies & People
- Predicting AGI by tracking top lab alumni migration patterns — dosco · 2026-09-16
- Wozniak returns to X with merch launch; Scoble recalls being his first customer — tbirdcymru · 2026-09-16
- Hiroshi Ishii's MIT Tangible Interfaces course draws 70+ students, 80% from Harvard — ishii_mit · 2026-09-16
- Anthropic's Head of Product Design Names Only Two Outside Tools It Uses — soleio · 2026-09-16
- Scoble on Zuck's Image Turnaround: Six Hours Walking With Zuckerberg — Scobleizer · 2026-09-16
- AI Eval Project Seeks Post-PhD Biology Experts as Human Graders — iskander · 2026-09-16