The Biology of Language: Why Artificial Intelligence May Become the Greatest Proliferation Challenge of the Twenty-First Century

By Matthew Parish, Associate Editor

Tuesday 9 June 2026

A remarkable joint letter sent to the United States Congress last week by the leaders of the world’s principal artificial intelligence companies marks an important moment in the evolution of the debate surrounding advanced machine learning systems. Executives including Sam Altman, Dario Amodei, Demis Hassabis and Mustafa Suleyman have called upon lawmakers to strengthen controls over the synthesis of DNA and RNA, warning that increasingly capable artificial intelligence systems may substantially reduce the expertise required to create dangerous biological agents.

The significance of this intervention should not be underestimated. For years, public discussion about artificial intelligence has tended to focus upon dramatic visions of autonomous machines overthrowing humanity, replacing all employment or somehow acquiring independent consciousness. Such scenarios attract headlines because they belong as much to science fiction as to public policy. Yet the concern expressed in the letter is considerably more immediate and arguably more credible. It concerns not machines acting alone, but machines amplifying the capabilities of human beings.

Biological weapons have traditionally been constrained by a number of barriers. Expertise in molecular biology is difficult to acquire. Laboratory equipment is expensive. The design of novel pathogens requires extensive knowledge accumulated over many years. Even state-sponsored biological warfare programmes have historically struggled with the complexity of manipulating living organisms.

Artificial intelligence threatens to erode one of those barriers: knowledge itself.

Large language models are, at their essence, systems for organising and retrieving information. They can synthesise vast quantities of scientific literature, explain technical concepts in accessible language and assist researchers in solving complex problems. These capabilities are precisely what make them useful for legitimate scientific research. They are also what makes them potentially dangerous.

A person who previously lacked advanced biological training can now ask questions in ordinary language and receive explanations that might once have required months of study. Researchers have repeatedly demonstrated that advanced language models can assist users in understanding sophisticated scientific subjects far beyond the user’s initial level of expertise. The concern is not that a machine independently invents a biological weapon. Rather, it is that a determined human actor gains access to an extraordinarily patient, knowledgeable and inexpensive tutor.

This phenomenon is not unique to biology. Artificial intelligence lowers informational barriers across virtually every discipline. A student can learn advanced mathematics more rapidly. An engineer can write software more efficiently. A lawyer can conduct research more effectively. Most of the time these developments are beneficial.

However biology possesses a unique characteristic that distinguishes it from many other sciences. Living organisms replicate themselves. A biological agent released into the environment may spread far beyond its original target. A poorly designed computer program can cause inconvenience. A poorly designed pathogen can potentially affect entire populations.

Historically the principal weapons of mass destruction have required substantial industrial infrastructure. Nuclear weapons demand enrichment facilities, specialised materials and highly visible engineering programmes. Chemical weapons require industrial-scale production. Biological weapons have always occupied a more ambiguous position because living organisms reproduce naturally. As biotechnology becomes cheaper and more accessible, the distinction between a sophisticated state programme and a smaller-scale operation may become increasingly blurred.

The AI executives’ letter therefore focuses not upon artificial intelligence alone but upon the intersection of artificial intelligence and synthetic biology. They advocate mandatory screening of synthetic genetic material orders and improved record-keeping throughout the gene synthesis industry. Their argument is that while artificial intelligence may make the design process easier, governments can still establish controls over the physical acquisition of genetic materials.

This is an important recognition that technological risks rarely emerge from a single invention. Catastrophic capabilities often arise when multiple technologies converge. The First World War did not result merely from the invention of explosives, but from the combination of industrial manufacturing, railway logistics, telecommunications and modern chemistry. Nuclear weapons emerged from the convergence of theoretical physics, industrial engineering and wartime state mobilisation.

Today the convergence occurs between artificial intelligence, cloud computing, automated laboratory equipment and synthetic biology.

The challenge for policymakers is that information itself is difficult to regulate. Liberal democracies are understandably reluctant to restrict scientific knowledge. Much of the biological information that might potentially be misused is already available in academic journals and public databases. Artificial intelligence does not create this information; it organises and explains it.

Consequently policymakers face a dilemma. Excessive restrictions may inhibit legitimate scientific research and medical innovation. Insufficient safeguards may create opportunities for malicious actors. The objective therefore cannot be the elimination of risk. Rather it must be the management of risk through layers of protection.

One of the most striking aspects of the letter is the unusual consensus it represents. The artificial intelligence industry has become deeply divided over questions of regulation, competition and safety. Rival companies frequently disagree about how rapidly technology should advance and how governments should respond. Yet on the issue of biological weapons there appears to be an emerging consensus that the threat deserves urgent attention.

This consensus is itself informative. The executives involved possess perhaps the most detailed understanding available of the capabilities and limitations of frontier language models. They are not warning that artificial intelligence has already enabled biological catastrophe. They are warning that the trajectory of capability improvement suggests that existing safeguards may soon become inadequate.

Whether one accepts every prediction made by artificial intelligence companies is another matter entirely. Critics reasonably observe that technology firms possess incentives to emphasise the transformative nature of their own products. Some previous warnings about artificial intelligence have been criticised as exaggerated or speculative. Yet even sceptics should recognise that biological security occupies a different category from many other artificial intelligence debates. When the potential consequences include large-scale loss of life, prudence becomes rational even under conditions of uncertainty.

For Ukraine, these questions are not abstract. The country has spent years confronting unconventional threats, disinformation campaigns and novel forms of warfare. The experience of war has demonstrated repeatedly that technological innovation can alter the balance of power far more rapidly than political institutions can adapt. Drones, satellite communications and machine-learning systems have already transformed the battlefield. It would be naive to assume that biotechnology will remain insulated from similar pressures.

The broader lesson is that artificial intelligence is not merely a technology. It is a force multiplier. It enhances human capabilities across almost every intellectual domain. In medicine this may produce extraordinary advances in diagnosis and treatment. In science it may accelerate discovery. In education it may democratise access to knowledge.

Yet every force multiplier carries corresponding risks. The same systems that help a researcher discover a cure may help a malicious actor understand dangerous biological concepts more quickly than ever before.

The danger therefore lies not in the machine itself, but in the unprecedented accessibility of expertise. Humanity has entered an era in which specialised knowledge can increasingly be delivered through ordinary conversation. The democratisation of knowledge may become one of the greatest achievements of artificial intelligence. It may also become one of its greatest security challenges.

The letter to Congress should therefore be understood not as an alarmist prediction of imminent catastrophe, but as an acknowledgement of a fundamental reality. The barriers that once separated ordinary individuals from highly specialised scientific knowledge are weakening. As those barriers fall, societies must develop new forms of governance capable of preserving the benefits of scientific progress while reducing the opportunities for catastrophic misuse.

That challenge may prove to be one of the defining policy questions of the twenty-first century.