AI and Biosecurity: Why Scientists Are Asking Harder Questions

Artificial intelligence has become part of how new medicines and vaccines are designed. Models trained on chemistry and on protein sequences help researchers narrow millions of possibilities down to a few worth testing in a laboratory. The benefits are real: faster work on rare diseases, new antibiotics, better vaccines, enzymes that break down plastic.

Some of the scientists building those tools have also raised a harder question in public: what happens if the same capabilities are misused? That question, usually called dual-use risk, is now a mainstream topic in biosecurity rather than a fringe worry.

Who's raising it

Dr. Sean Ekins, a pharmacologist whose company uses machine learning to look for treatments for rare and infectious diseases, co-authored a 2022 paper in Nature Machine Intelligence on the dual-use potential of AI-driven drug discovery, written for a Swiss conference on chemical and biological arms control. In 2023, he and colleagues wrote in the Bulletin of the Atomic Scientists about protein design models, which they described as a kind of "ChatGPT for biology."

One of his co-authors, Dr. Filippa Lentzos, is a Reader in Science and International Security at King's College London. She chairs the World Health Organization's technical advisory group on the responsible use of the life sciences and dual-use research, and is on the UN roster of experts for investigating alleged use of chemical or biological weapons. Her point is institutional rather than technical: scientific capability doesn't stay with the team that developed it. Papers are read everywhere, software is copied, and services make laboratory work available to people who didn't build the tools.

Why detection is part of the worry

Much of biodefense works by recognition. Laboratories and screening systems look for known genetic sequences, known toxins and known chemical signatures. A substance that has a name can be put on a control list. The concern raised by researchers in this field is that systems which look only for what has already been catalogued may be a weaker safeguard over time than people assume, which is one reason the WHO, national AI safety institutes and the companies developing these models have all taken up the question of safeguards.

What it does and doesn't mean for a household

None of this means an attack is coming, and families don't need to follow the science to make sensible decisions. What the debate does suggest is that biological and chemical risks aren't a closed chapter from the twentieth century; serious people in medicine, security and public health are treating them as live questions.

For a household, the practical response is the same as for any airborne hazard, and it's worth doing before any warning arrives:

  • Know where you'd shelter and keep it ready. See our guide to setting up a safe room.
  • Keep enough water, food and medication for several days at home.
  • Have a mask or hood that fits each person, with sealed spare filters: the Israeli 4A1 for clean-shaven adults, the Sapphire hood for beards and glasses, and powered hoods or youth masks for children.
  • Follow public health instructions in an outbreak, which matter far more than equipment.

Our guide to biological threats and respiratory protection explains what a filter does and doesn't do against germs, and Andy Weber on biodefense covers the policy side.

Frequently asked questions

Is AI making biological weapons easy to make?

No. Making and delivering a biological or chemical weapon remains difficult and heavily monitored. The concern researchers raise is about the long-term direction of the risk, not about something anyone can do today.

Who's responsible for managing this?

Governments, the WHO, arms-control bodies, national AI safety institutes and the companies developing the models, working with scientists in the field.

Does any of this change what a family should buy?

No. The same shelter plan, masks and supplies cover biological, chemical and radiological hazards.

About this article: it refers to published work by Sean Ekins, Filippa Lentzos and colleagues in Nature Machine Intelligence (2022) and the Bulletin of the Atomic Scientists (2023), and to the WHO's work on dual-use research. Those researchers and organizations are not affiliated with CBRNMASKS.COM and have not endorsed it or its products. Analysis and recommendations are by David Magen, founder of CBRNMASKS.COM. Former combat research officer in the IDF Training and Doctrine Branch and former reserve officer in Israel's National Emergency Authority.

Primary Sources

  • Fabio Urbina, Filippa Lentzos, Cédric Invernizzi, Sean Ekins — "Dual Use of Artificial-Intelligence-Powered Drug Discovery," Nature Machine Intelligence, March 2022
  • Sean Ekins, Filippa Lentzos, Max Brackmann, Cédric Invernizzi — "There's a 'ChatGPT' for Biology. What Could Go Wrong?" Bulletin of the Atomic Scientists, March 2023
  • WHO — Responsible Use of the Life Sciences and Dual-Use Research
  • UK AI Security Institute — Frontier AI Trends 2025

Analysis and preparedness conclusions by David Magen, founder of CBRNMASKS.COM; former combat research officer in the IDF Training and Doctrine Branch; former reserve officer in Israel's National Emergency Authority. Sean Ekins, Filippa Lentzos, Nature Machine Intelligence, Bulletin of the Atomic Scientists, and the WHO are not affiliated with CBRNMASKS.COM and have not endorsed the company or its products.

Back to blog