
Scientists have used artificial intelligence to generate entirely new viruses for the first time, training an AI model on DNA sequences much as ChatGPT learns from text.
The bacteriophages created are harmless to humans and represent an advance in biological research, but the technology presents a dual-use risk: the same capabilities could enable less-skilled actors to design dangerous pathogens more easily than ever before.
What happened
Researchers have unveiled the first viruses generated entirely by artificial intelligence, using a model trained on DNA sequences (similar to how ChatGPT learns from text) to design new genomes with minimal human oversight. The bacteriophages created are harmless to humans.
Why it matters
While engineered viruses have long been used safely in genetic therapies, AI generation lowers the technical barrier to creating them. The same capabilities that enable benign research could be repurposed to design dangerous pathogens, potentially within reach of people lacking deep biological expertise.
What to watch
The tension between AI's research benefits and dual-use risks. An AI safety campaigner has framed the problem as "Moore's law of mad science": every 18 months, the minimum skill required to cause biological harm decreases by one point — a warning that broader access to these tools may outpace safety guardrails.
Researchers have unveiled the first viruses created entirely through artificial intelligence, a milestone that both demonstrates the power of AI-driven biology and crystallizes long-standing concerns about dual-use technology.
The scientists trained an AI model on DNA sequences—much in the way ChatGPT learns patterns from text—and used it to generate novel viral genomes with far less human input than traditional approaches. The viruses they created are bacteriophages, which are harmless to humans and have legitimate research applications. This is not the first engineered virus: viral vector therapies have long used carefully designed viruses to carry genetic code into cells for medical purposes. But the new approach is novel in its automation and the minimal human guidance required to produce working designs.
The breakthrough carries a darker implication. Because the same AI capabilities that enable safe bacteriophage research could theoretically be redirected toward pathogenic viruses, and because the technology lowers the technical bar for deploying it, the research underscores what the article calls the "double-edged sword" nature of biological tools. An AI safety advocate quoted in the piece coined the phrase "Moore's law of mad science": "Every 18 months, the minimum IQ to destroy the world drops by one point." The remark encapsulates the risk that as AI systems become more powerful and accessible, the skills and resources needed for a malicious actor to design a dangerous pathogen will continue to fall, potentially outpacing the development of regulatory safeguards.
The creation of AI-generated viruses marks a milestone in computational biology, extending the pattern of AI systems trained on biological data (here, DNA sequences) to design novel organisms. The work builds on established knowledge—engineered viruses have been a staple of genetic medicine for years—but the automation and reduced human input represent a qualitative shift in accessibility.
The body itself identifies the core concern: dual-use risk. Biological tools, the article notes, are inherently "double-edged swords." A model that can generate harmless bacteriophages for research can, in principle, be adapted to design pathogens. The safety risk is not that the technology is new in concept (custom viruses are not), but that AI dramatically reduces the expertise and resources a person would need to deploy it. An AI safety campaigner quoted in the article frames this as "Moore's law of mad science"—the idea that every 18 months, the minimum IQ (or by extension, skill and resources) to cause biological destruction drops by one point. This suggests a race between capability expansion and safety norm-setting, with the former outpacing the latter.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
Amazon and Google are intensifying competitive efforts against The Trade Desk (TTD), a major digital advertisi…
OpenAI introduced Premium Seats for ChatGPT Business, priced at $125 per user per month ($100 with annual bill…

Computer scientists at University of Tübingen, Max Planck Institute, MATS Research, and Snyk discovered a meth…

Anthropic pledged to embed machine-readable watermarks in Claude-generated text and digitally signed provenanc…

Anthropic has signed the EU AI Act Code of Practice and will embed invisible watermarks in Claude-generated te…

Meta CEO Mark Zuckerberg published a 6,500-word essay Monday outlining his vision for artificial intelligence…

The AI news that matters, in one minute each morning.
Sign up free