increasing the severity of future terrorist attacks. One area of particular concern is dual use of AI models: In a laboratory experiment reported in 2022, an AI system that had previously been used for medicinal drug discovery was trained to find and combine toxicity molecules. Within only six hours, 40,000 compounds at least as toxic as the sample nerve agent had been generated. 55 Theoretically, there is an unlimited number of new toxic substances that could be created using such models. 56 The researchers involved in the experiment emphasized that the computing power and software required for such experiments is easily attainable today. 57 Experts are also warning about the relative ease with which viruses capable of infecting humans, such as monkeypox or smallpox, could be enhanced to evade human immune systems, making standard vaccines ineffective. With the tools and information required to alter a pathogen’s genetic code becoming easier to access, it may only be a matter of time before a threat actor releases a virus that causes the next pandemic. 58 As the costs of setting up a laboratory and purchasing the necessary equipment are relatively limited, the main barrier to threat actors misusing advances in biotech is having the scientific expertise itself – a barrier that will be far from insurmountable over the next decade. Of course, it will also take considerable (and unrelated) expertise to translate the creation of new toxic substances into the building of weaponry, given the complexities of transporting and disseminating the substances created. But unlike in the nuclear sector, where strict protocols and monitoring of materials and equipment make proliferation efforts relatively easy for governments to detect, this set of conditions is not present in the same way when it comes to weaponizing biotech. Biotech can also provide a bridge from the biological world to the cyber realm. As far back as 2017, researchers in the United States demonstrated that it was possible to hack a computer using DNA sequence data. Under certain preconditions, they were able to introduce malware into DNA purchased online (at minimal cost), which was read and then processed by a computer that in turn became compromised by the malware. 59 Looking ahead a decade, as Cyber espionage and warfare becomes more sophisticated and more people become acquainted with biotech developments, it is conceivable that the researchers’ warning – that hackers could use the DNA sequences from faked blood samples to gain access to hack computers – could come true. 60 Indeed, GRPS respondents express concern with the risk interconnectionbetween Adverse outcomes of frontier technologies and Cyber espionage and warfare,as shown in Figure 2.11 below. Adverse outcomes of AI technologies Censorship and surveillance Crime and illicit economic activity Cyber espionage and warfare Decline in health and well-being Disruptions to a systemically important supply chain Inequality Lack of economic opportunity or unemployment Misinformation and disinformation Online harms Societal polarization Adverse outcomes of AI technologies Censorship and surveillance Crime and illicit economic activity Cyber espionage and warfare Decline in health and well-being Disruptions to a systemically important supply chain Inequality Lack of economic opportunity or unemployment Misinformation and disinformation Online harms Societal polarization Adverse outcomes of frontier technologies Adverse outcomes of frontier technologies Biological, chemical or nuclear weapons or hazards Biological, chemical or nuclear weapons or hazards Risk interconnections: Adverse outcomes of frontier technologies FIGURE 2.11 Source World Economic Forum Global Risks Perception Survey 2024-2025.Edges Relative inﬂuence High LowMediumRisk inﬂuenceNodes High LowMediumReferenceRisk categories Economic Environmental Geopolitical Societal Technological Global Risks Report 2025 56