Back to the MIT repository
4. Malicious Actors & Misuse2 - Post-deployment

Biological and chemical attacks

Growing evidence shows general- purpose AI advances beneficial to science while also lowering some barriers to chemical and biological weapons development for both novices and experts. New language models can generate step- by- step technical instructions for creating pathogens and toxins that surpass plans written by experts with a PhD and surface information that experts struggle to find online, though their practical utility for novices remains uncertain. Other models demonstrate capabilities in engineering enhanced proteins and analysing which candidate pathogens or toxins are most harmful. Experts could potentially use these in developing both more advanced weapons and defensive measures.

Source: MIT AI Risk Repositorymit1023

ENTITY

1 - Human

INTENT

1 - Intentional

TIMING

2 - Post-deployment

Risk ID

mit1023

Domain lineage

4. Malicious Actors & Misuse

223 mapped risks

4.2 > Cyberattacks, weapon development or use, and mass harm

Mitigation strategy

1. Impose stringent access controls and technical safeguards on advanced AI models possessing biological or chemical dual-use capabilities, including the restriction of model weights and the removal of hazardous functionalities from general-purpose systems. 2. Establish a standardized, AI-enabled nucleic acid synthesis-screening protocol and enhance regulatory oversight for cloud laboratories and genetic synthesis providers to detect and prevent the illicit procurement of high-risk biological agents. 3. Mandate comprehensive biosecurity risk assessments and integrate a dynamic governance framework that includes developer liability and continuous monitoring to ensure proactive, iterative adaptation to the rapidly evolving threat landscape.