AI can design viruses, toxins and other bioweapons. How worried should we be?

2026-05-13 · Source: Machine learning : nature.com subject feeds · Field: Science & Research — Life Sciences & Biology, Health & Medical Research, Biosecurity · Depth: Intermediate, long

Summary

The development of an AI tool by Chinese scientists in 2024 to design conotoxins, a class of small proteins found in venomous cone snails, has raised biosecurity concerns among some US government employees and researchers. While the Chinese team states the tool is for drug discovery, with some designs showing therapeutic potential, the incident highlights growing anxieties about AI's dual-use potential in biology. Experts debate whether AI could facilitate the creation of novel bioweapons, such as undetectable toxins or modified pandemic viruses like SARS-CoV-2. A 2025 NASEM report suggests significant barriers exist for AI to enhance pandemic pathogens, primarily due to data limitations and lab production difficulties. However, the report acknowledges AI could design toxins for targeted attacks, and studies show AI can increase the expertise of bad actors, potentially circumventing existing DNA synthesis screening protocols.

Key takeaway

For CTOs and executives overseeing biotechnology or AI development, you must prioritize robust biosecurity measures and ethical AI deployment. The demonstrated ability of AI to design novel toxins and circumvent DNA synthesis screening, even if challenging to implement physically, necessitates proactive risk assessment. You should invest in advanced detection and countermeasure technologies, advocate for mandatory global screening standards for nucleic acid synthesis, and implement stringent internal guard rails on biological AI models, acknowledging their potential for circumvention.

Key insights

AI tools in biology present a dual-use dilemma, offering drug discovery benefits while raising bioweapon design and proliferation risks.

Principles

• AI can enhance bad actors' biological expertise.
• Guard rails on AI models can be circumvented.
• DNA synthesis screening is a critical bulwark.

Method

AI models can design novel proteins and modify existing biological sequences, potentially enhancing pathogenic properties or creating new threats, with some success demonstrated in lab settings.

In practice

• AI can generate code for lab robots.
• AI can troubleshoot virology protocols.
• AI can design synthetic homologues to evade screening.

Topics

• AI Biosecurity
• Biological AI Tools
• Toxin Design
• DNA Synthesis Screening
• Large Language Models

Best for: CTO, Executive, AI Scientist, Research Scientist, Policy Maker

Related on AIssential

• Using synthetic biology and AI to address global antimicrobial resistance threat — Synthetic biology and AI can develop programmable antibacterials to combat rising antimicrobial resistance.
• An illegal bioweapons lab was found in a Las Vegas garage. It’s a warning for Australia — AI and open-source biological data are enabling the proliferation of illegal biolabs and the potential for bioweapon creation.
• Job titles of the future: Nature’s drug designer — Precision drug design, enhanced by AI, can be applied to conservation for targeted ecological health solutions.
• Radical Numerics launches with $50M to build general biological intelligence — Multimodal AI models integrating diverse biological data can advance human health through diagnostics and drug discovery while simultaneously bolstering biosecurity.
• AI can design and run thousands of lab experiments without human hands. Humanity isn’t ready for the new risks this brings to biology — AI is rapidly automating biological experimentation, accelerating discovery but also creating significant dual-use biosecurity risks.
• The Dog That Hacked Cancer — AI tools can accelerate personalized medicine development, even for non-experts.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine learning : nature.com subject feeds.