The Dual Potential of AI in Pandemics: Sparking or Saving Lives.

Large language models (LLMs) like ChatGPT can provide detailed instructions on how to carry out a biological weapons attack or build a bomb, making it easier for non-experts to access such information. While Google and other search engines have made it difficult to learn how to commit acts of terrorism, LLMs pose a new threat. Experts suggest better controls at all chokepoints, including requiring DNA synthesis companies to screen orders and removing papers about dangerous viruses from training data for powerful AI systems. Positive actors in the biotech world are taking this threat seriously, with Ginkgo Bioworks partnering with US intelligence agencies to develop software that can detect engineered DNA at scale.