ADAPT: a programme for the advanced detection of AI-enabled pathogenic threats

Area of Science:

• Biosecurity
• Synthetic Biology
• Bioinformatics

Background:

• Artificial intelligence (AI) advances biological engineering, creating novel biosecurity threats.
• Existing synthetic nucleic acid screening methods are insufficient for AI-generated threats.
• The Functional Genomic and Computational Assessment of Threats (FunGCAT) program improved sequence screening but left critical gaps.

Purpose of the Study:

• To propose the Advanced Detection of AI-enabled Pathogenic Threats (ADAPT) program as a successor to FunGCAT.
• To address limitations in defining and detecting AI-driven biological threats.
• To develop next-generation screening tools for biosecurity.

Main Methods:

• Phase I: Develop a function-based definition for sequences of concern and create benchmark datasets.
• Phase II: Develop and validate screening tools for known threats, AI-functional homologues, and novel AI-designed threats.
• Integrate continuous governance for regulatory translation and secure infrastructure.

Main Results:

• The study outlines a two-phase program (ADAPT) to bridge current detection gaps.
• It proposes a shift from sequence similarity to function-based threat detection.
• It aims to create tools capable of identifying AI-generated biological threats.

Conclusions:

• ADAPT is crucial for adapting biosecurity screening to the era of AI-driven biological engineering.
• The program will enhance detection capabilities against novel and evolving biological threats.
• It builds upon previous efforts while addressing the unique challenges posed by AI in synthetic biology.