War on the Rocks: Security Pros Say Aum Failure Shows Bioterror Risk Overstated; Industry Warns of AI
Law enforcement and security professionals say Aum Shinrikyo’s failed biological weapons program in the early 1990s shows the risk of bioterrorism is overstated, even as industry executives and researchers keep warning about artificial intelligence and calling for better preparedness, War on the Rocks reported.
Why did Aum Shinrikyo fail?
The Japanese doomsday cult responsible for the 1995 Tokyo sarin attack possessed financial resources, facilities, scientifically educated members, and leadership willing to cause mass harm. Despite these advantages, the group’s biological weapons program repeatedly failed. Biological production involves multiple interdependent problems like acquisition, cultivation, scaling, purification, weaponization, and deployment.
Aum encountered real scientific barriers during its attempts. However, structural and organizational dynamics made an already difficult undertaking harder by preventing knowledge from directing key decisions. Secrecy restricted access to relevant expertise, while hierarchy placed authority in the hands of leaders lacking technical judgment.
What were the expertise gaps?
Given his relevant academic background, Seiichi Endo was appointed to lead Aum’s biological weapons initiative, yet proficiency in biology is not a monolithic capability. Endo studied veterinary medicine and started a graduate degree in virology, but concentrated primarily on bacterial pathogens and their toxins. His background lacked in-depth, pathogen-specific knowledge of the agents the cult tried to weaponize, particularly concerning a virulent strain of Bacillus anthracis.

Complex engineering efforts require interdisciplinary knowledge spanning microbiology, aerobiology, chemistry, statistics, and mechanical engineering. Endo delegated critical tasks to members with zero biology training. When Asahara ordered members to construct a commercial spraying system due to anticipated delays, the resulting machine broke down repeatedly.
How does artificial intelligence change risks?
AI has the potential to address both technical and organizational hurdles that contributed to past failures. Extremists are already exploring these tools, as seen when Boko Haram factions built specialized AI units to route difficult questions and supplement knowledge, according to research cited by War on the Rocks.
General-purpose language models can troubleshoot virology protocols and help design specialized tools. Guided by prompts, models can design potential toxin binders and yield successful laboratory results. Recent models such as Pinal take textual descriptions of desired functions and output protein designs.
What does laboratory testing show?
Existing studies provide mixed evidence regarding real-world uplift from AI assistance. A small pilot study with participants from the Los Alamos National Laboratory saw novices succeed more often with AI access, though a smaller Anthropic study found no differences.
Meanwhile, the U.K. AI Security Institute reports that their internal investigations demonstrate how beginners can achieve success with difficult wet lab procedures when utilizing a language model. Developers like Anthropic and OpenAI have implemented safeguards, but open-weight models and third-party services continue to present challenges to restricting access.
Can security safeguards prevent misuse?
While security measures can mitigate the risk of abuse, the effectiveness of these protections differs between organizations, and external researchers have discovered ways to bypass the guardrails of frontier models. Open-weight models allow users to download and host systems independently on consumer hardware.
Aum Shinrikyo evolved from a yoga school founded in 1984, officially gaining status as a religious organization in 1989. Known as a “religion for the elite,” the group attracted a significant number of graduates from Japan’s top universities and became one of the fastest-growing religious movements in the country’s history.
Its early public outreach cleverly utilized anime and manga themes to integrate religious ideas with science fiction narratives, including the quest for ultimate truth and the use of powerful weapons. Despite its rapid growth and institutional backing, the organization was later formally designated as a terrorist group by several countries, including Japan, Russia, Canada, and members of the European Union, following its involvement in multiple sarin attacks.
Users can remove built-in safeguards through fine-tuning, placing use beyond developer monitoring. Consequently, security analysts argue that past failures of well-resourced non-state groups can no longer serve as proof that future organizations will fail.
Related reading
- Indonesia Elected to ICAO Council, Ending 25-Year Absence
- U.S. Startup Sues Apple in India Over NameDrop Patent, Seeks ₹15 Billion and Import Ban
- US Panel Warns China Seeks to Contain India With Pakistan Partnership (time.news)
- 3.6% for Social Security, 2.6% for FERS, and a Bill Still Pending for VA (daybreakwire.com)