Researchers Refute AI Doomsday Scenarios, Citing Gap Between Digital Models and Physical Harm

Multiple AI researchers, including Google's Anselm Levskaya and MBZUAI's Eric Xing, argue that extreme AI doomsday scenarios are heavily exaggerated. They emphasize that while AI can generate digital virus blueprints, translating digital outputs into real-world harm requires physical equipment, wet-lab synthesis, and extensive biological testing. This perspective shifts the debate from sci-fi existential doom toward realistic tech policy and safety priorities. It highlights that current AI capabilities remain bound by physical laws and material constraints, remaining primarily tools guided by human intentions rather than autonomous, self-directed threats. Experts including Cohere CEO Aidan Gomez and IDSIA Scientific Director Jürgen Schmidhuber stressed the importance of distinguishing between human misuse of AI and autonomous AI self-direction. However, law professor Mateusz Błaszczyk cautioned that debunking doom scenarios should not lead to ignoring realistic, immediate harms like AI-driven cyberattacks and surveillance.