# OpenAI Develops Framework to Assess AI's Role in Biological Threat Creation

OpenAI announced it is building an early warning system to evaluate whether large language models could help bad actors create biological threats.

In a first-of-its-kind study, the company tested GPT-4 with both biology experts and students to measure whether the AI provided meaningful assistance in biological threat creation. The results showed GPT-4 offered only a "mild uplift" in accuracy—not enough to draw definitive conclusions but significant enough to warrant attention.

This research represents OpenAI's attempt to get ahead of potential misuse as AI systems become more capable. The company is creating a blueprint that other organizations can use to assess similar risks in their own models.

**Why it matters:** As AI becomes more powerful and accessible, understanding whether these tools could enable dangerous activities is critical for safety. OpenAI's framework could become an industry standard for evaluating