Back
UNICRI: Dual-Use Research and Artificial Intelligence
reportPublished by UNICRI (UN Interregional Crime and Justice Research Institute) in 2021, this report is relevant to AI governance discussions around how powerful AI tools in biology may require international oversight to prevent catastrophic misuse.
Metadata
Importance: 58/100organizational reportanalysis
Summary
This UNICRI report examines the dual-use risks of AI-driven protein-folding prediction tools like AlphaFold, which offer major benefits for medicine but could be weaponized to engineer pathogens or target specific populations. It calls for governance frameworks that balance scientific openness with biosecurity safeguards.
Key Points
- •AlphaFold and similar AI tools accelerate drug discovery and vaccine development but also lower barriers to identifying disease-causing protein mutations.
- •Malicious actors could exploit protein-folding AI to engineer more virulent pathogens or identify population-specific biological vulnerabilities.
- •The paper frames this as a classic dual-use dilemma requiring proactive governance rather than reactive regulation.
- •Authors advocate for international coordination mechanisms to monitor and restrict dangerous applications of AI-enabled biotechnology.
- •Highlights a broader pattern where AI capabilities amplify both beneficial and harmful potential in sensitive scientific domains.
Cited by 1 page
| Page | Type | Quality |
|---|---|---|
| Bioweapons Risk | Risk | 91.0 |
Cached Content Preview
HTTP 200Fetched Mar 20, 20268 KB
# THE POTENTIAL FOR DUAL-USE OF PROTEIN-FOLDING PREDICTION
by Sterling Sawaya, Taner Kuru, Thomas A. Campbell,
A rtificial intelligence (AI) offers great potential for scientific advancement, particularly in areas with high complexity and many variables. Although biology can be challenging to comprehend, scientists can develop accurate AI models that help resolve biological interactions more effectively than conventional methods. Implementing AI in biological sciences promises benefits for all human beings by improving our understanding of how molecules interact or how genetic variation influences traits.
Traditional protein prediction methods rely on time-intensive methods requiring expensive equipment. Unfortu
66
# AI offers the potential to rapidly and cheaply predict protein structures, allowing scientists to examine the impact of genetic variations on macromolecules
nately, it is difficult to predict some complex protein classes. AI offers the potential to rapidly and cheaply predict protein structures, allowing scientists to examine the impact of genetic variations on macromolecules. AI-based predictions are expected to improve our understanding of how genetic variation causes disease, enables drug development, and allows researchers to design specific proteins that break down plastics in nature. Consequently, protein-folding prediction is a significantly important domain for the implementation of AI in biological sciences.
Recently, exponentially rapid implementation and advancement of AI in protein-folding prediction have yielded surprising results. In 2020, the company DeepMind made a major step forward with its algorithm AlphaFold at the 14th Community Wide Experiment on the Critical Assessment of Techniques for Protein Structure Prediction (CASP). AlphaFold predicted structures for a range of proteins with an accuracy over $9 0 % ,$ , comparable to conventional methods.1 From this success, highly accurate protein structure prediction may soon promise benefits for public health, such as vaccine developments.
However, malicious actors might use protein-folding prediction algorithms for purposes unintended in their original development. Understanding how genetic variation affects protein structure may enable researchers to discover disease-causing mutations using one individual’s genome sequence. While this ability
# However, malicious actors might use protein-folding prediction algorithms for purposes unintended in their original development
might raise serious concerns for individuals who carry rare and/or severe genetic diseases, the threat would become even more significant if malicious actors could predict how genetic variations impact the severity of diseases and identi
... (truncated, 8 KB total)Resource ID:
80cf8f51eecba79e | Stable ID: ZmNjMjI4Ym