BlueDot Narrated AI Is Reviving Fears Around Bioterrorism. What’s the Real Risk?
Sep 12, 2025
The conversation dives into the unsettling prospect of AI being used for bioterrorism, raising fears as large language models proliferate. Experts highlight instances where chatbots provided guidance on weaponization, revealing a chilling increase in technological capabilities. Barriers to creating bioweapons are lowering, making it easier for rogue actors to exploit these advances. The discussion also covers the dual nature of AI, capable of facilitating attacks while also developing countermeasures, indicating a complex future ahead.
AI Snips
Chapters
Transcript
Episode notes
LLMs Are Making Technical Gaps Smaller
- The proliferation of LLMs is increasing security risks by making technical guidance on bioweapons easier to access.
- Assessments show models' ability to instruct on acquiring and releasing lethal agents improved sharply in 2024.
White House Demo With Chatbot-Recommended Precursors
- In early 2023 Rocco Casagrande brought a container of common chemicals to the White House after a chatbot recommended them as pandemic precursors.
- The stunt illustrated how an internet connection and an LLM could enable dangerous experimentation.
Conflicting Assessments Mask Fast Change
- Not all studies agree on the magnitude of AI's advantage; OpenAI found GPT-4 gave only marginal benefits over internet searches in 2023.
- Yet other reports show rapid year-on-year improvements, so risk is evolving, not fixed.