Researchers argue AI-generated virus threat overstated, highlight real-world constraints
Experts say the notion that AI could instantly create a deadly virus ignores the practical steps needed to produce and test a biological agent.
The debate over AI-driven existential risk has intensified after Jacob Coxon left Anthropic and warned that companies like OpenAI are racing toward self-improving superintelligence. Although Evan Hubinger estimates a greater-than-10% chance of AI wiping out humanity within a decade, many researchers stress the distinction between digital design and real-world execution. Anselm Levskaya and Eric Xing explain that turning a computer-generated viral sequence into an actual virus demands laboratory synthesis, particle production, and extensive testing—processes that cannot be automated by AI alone.
Oleksandr Yaremchuk adds that any harmful AI action would still need access to physical resources, money, or people. Commentators such as Aidan Gomez and Jürgen Schmidhuber argue that most current risks stem from humans using AI as tools, not from autonomous AI scientists, while legal scholar Mateusz Blaszczyk cautions against ignoring nearer-term AI threats like cyber-attacks and surveillance.
