AI extinction risk and bioweapons: are we ready?

Overview

The double-edged sword effect of artificial intelligence technology is provoking deep concern among global security experts. Recently, a roundtable hosted by MIT Technology Review focused on whether AI possesses the potential to "extinguish humanity," with AI-assisted bioweapons development capability emerging as the most prominent focal point. Meanwhile, concerns raised by Microsoft and OpenAI employees about AI eroding the online content ecosystem, alongside geopolitical maneuvering among nations over scientific and technological self-reliance, collectively paint a complex and severe picture of the current state of this technical field.

AI Biosecurity: The Perilous Leap from Laboratory to Battlefield

Despite the breakthrough advancements AI has made in drug discovery, its potential catastrophic consequences have begun to surface. A 2022 study revealed that an AI "molecular generator" designed for drug development could produce 40,000 molecular structures usable as chemical warfare agents in just six hours. This discovery stunned the scientific community, marking a significant lowering of the threshold for the malicious use of AI in designing lethal pathogens.

"This is the largest-scale theft of labor in human history." — Brent Hecht, Microsoft Applied Science Director, stated bluntly in court filings for the New York Times v. OpenAI copyright case.

Today, with the proliferation of gene-editing and synthetic biology tools, AI can answer questions across virtually all scientific domains. Although various countries have established certain safety guardrails, these measures are far from impenetrable. While scientists disagree on risk assessments, there is a consensus that the combination of biotechnology lacking ironclad defensive mechanisms with powerful AI poses a severe global security challenge.

Internal Rifts Among Tech Giants and the Crisis in Content Ecosystems

In the pursuit of computing power and model capabilities, tech giants have also experienced fierce internal conflicts over values and sustainability. According to reports from 404 Media and other outlets, employees at Microsoft and OpenAI have publicly stated that AI is "destroying the internet." These allegations primarily stem from AI crawlers scraping massive amounts of website data for training, causing the traditional publisher-dependent content business model on which they rely to teeter on the brink of collapse.

In the New York Times copyright lawsuit, these employee testimonies became crucial evidence, potentially further weakening the legal defense positions of OpenAI and Microsoft. Industry analysts point out that if the production of fresh content dries up due to lack of economic returns, the source of internet data upon which AI depends will also face depletion, ultimately undermining AI's own developmental foundations. This is not merely a legal issue, but a systemic risk concerning the survival of the information ecosystem.

Reshaping the Global Scientific Research Landscape and Responses

Faced with external technological restrictions, Chinese tech enterprises have demonstrated strong resilience and innovative vitality. The latest data shows that Sino-US tech friction has instead prompted Chinese companies to increase investment in basic science, with the number of related patent citations of scientific papers surging by 72% year-on-year. This "reverse incentive" effect indicates that technological containment has not halted innovation but has rather accelerated the autonomous progress of the local research system.

Meanwhile, the convergence of space exploration and robotics is also opening new battlefields. From the combat use of unmanned boats in Ukraine to humanoid combat robots developed by American companies, AI is rapidly penetrating the defense sector. Experts urge that while enjoying the dividends of technology, a stricter international regulatory framework must be established to ensure that AI remains under effective human control, preventing it from evolving from a "tool" into a "threat."