AI's bioweapon risks spark venture opportunities in biodefense screening and onshoring
AI models already outperform humans on most biology benchmarks since 2024, enabling non-state actors.
Austin Morrissey’s analysis on LessWrong connects the rapid progress of AI in biology to a looming biosecurity crisis. He notes that leading AI models have performed at or above human-expert level on most biology benchmarks since 2024, including tasks like designing DNA sequences for viral pathogens and driving protein-design tools. This capability diffusion to threat actors is inevitable, despite a common misconception that biological weapons are ethically taboo in modern conflict. Morrissey argues that for terrorist groups seeking unbounded mass harm, bioweapons are very much on the table. He uses the Aum Shinrikyo case as a historical precedent — a cult that attempted bioweapon attacks before AI existed. With AI now providing infinitely patient, cross-domain mentorship to aspiring bioterrorists, the likelihood of successful attacks increases. Even if only 0.1% of the global population harbors such intent, the absolute number of potential actors is substantial. The article calls for preparation for a 'Mythos-like moment' where national security fully wakes up to these risks.
Morrissey translates this risk into concrete venture opportunities. The most VC-shaped interventions fall into two categories: screening technologies (e.g., Know Your Customer systems for DNA synthesis orders) and onshoring manufacturing of chemical precursors used in both medicines and bombs. He points to the National Security Commission on Emerging Biotechnology as evidence of mounting political will. SecureBio, a nonprofit focused on catastrophic pandemic prevention, has already documented that AI can troubleshoot lab experiments and write code for lab automation — capabilities that dramatically reduce the expert knowledge barrier. For venture capitalists, these threats represent a neglected market opportunity. The article urges funding biodefense startups that can provide early-warning signals and resilient supply chains, rather than relying on ethical constraints that may not hold when adversaries prioritize mass harm.
- AI models have outperformed humans on most biology benchmarks since 2024, enabling non-state actors to design viral DNA and drive protein engineering.
- Historical case study of Aum Shinrikyo shows that even without AI, terrorist groups have attempted bioweapon attacks; AI now provides infinitely patient expert mentors.
- Venture opportunities exist in Know Your Customer screening for DNA synthesis and onshoring chemical precursor manufacturing, backed by political signals from the National Security Commission on Emerging Biotechnology.
Why It Matters
As AI lowers the barrier to bioweapons, venture-backed screening and onshoring solutions become critical for national security.