Russian Media: U.S. Military AI Genome Project May Have "Dual Use"

Reference News Network reported on May 8 An article published on May 5 on the website of the Russian Strategic Culture Foundation, titled "Pentagon Develops 'Absolute Biological Weapons'—Artificial Intelligence Unveils the Secrets of the Genome," authored by Vladimir Prokhvatilov. The full text is excerpted as follows:

Raytheon BBN Technologies has signed a contract numbered W911NF-17-2-0092 with the U.S. Army, which is part of the Department of Defense's Advanced Research Projects Agency's Functional Genomics and Threats Computational Assessment (FELIX) project. The FELIX project aims to develop methods for detecting genetic manipulation traces in biological systems to identify biological threats.

Dr. Jacob Beal of Raytheon BBN Technologies is a leading expert in developing genetic design methods using advanced artificial intelligence (AI) models. In the U.S., he is regarded as a key developer of specialized computer languages in the field of biological cells, helping to bridge the gap between "computer algorithms and biological cell programming."

Jacob Beal spent many years creating relevant computer languages; however, he suddenly "shifted direction" and began focusing on developing methods for "genomic threat detection," which can be used to discover covert biological weapons. In this new concept, deoxyribonucleic acid (DNA) is not viewed as a "mere living molecule," but as "compiled binary data streams." Genomic sequences will become a digital language that can be analyzed through computational models rather than studied in physical laboratories.

The project will apply a technology known as "Deep Packet Inspection" (DPI) to identify genomic threats. This is thanks to the contributions of Jacob Beal's collaborator, Daniel Wischgrod, a senior expert in U.S. cyber intelligence. Together, they introduced the DPI framework from the field of cybersecurity into genomics.

In cybersecurity, DPI technology detects malicious software code by analyzing the payload of network data packets; in genomics, DPI technology searches for "malicious intents" in genomic sequences that aim to bypass cellular defense mechanisms or produce harmful toxins by combining AI models.

Raytheon BBN Technologies' new research and development achievement is a "breakthrough in real-time biosafety." Jacob Beal and Daniel Wischgrod stated that this system is designed to guard against two specific scenarios: first, destructive activities conducted through legitimate biological laboratories, where malicious actors split the genomes of dangerous pathogens into tiny fragments, bypass government regulations to procure them in batches from commercial DNA synthesizers, and then assemble them into biological weapons; second, real-time detection of pathogens in samples to be tested, which is tactical biological reconnaissance in a combat environment.

In summary, the engineering logic of this patent is to achieve rapid detection of genomic threats.

At the military application level, according to the developers' vision, the military could create a digital radar for biological defense that directly integrates with portable gene sequencers. They could be deployed on unmanned reconnaissance aircraft of the three defense forces or installed in the ventilation systems of command bunkers.

The digital radar for detecting genomic threats must be able to instantly capture combat genetic payloads released by the enemy in smoke, issuing commands to initiate biological protection measures before personnel come into contact with pathogenic doses.

However, in addition to rapidly detecting genomic threats, this technology clearly has dual-use potential. By embedding Jacob Beal and Daniel Wischgrod's mathematical methods into a Generative Adversarial Network (GAN) architecture, it is possible to create "absolute biological weapons" that are invisible to monitoring systems.

GAN is an artificial intelligence deep learning architecture proposed by Ian Goodfellow in 2014, with two core networks as its main components. One is the generator, which generates new data (such as images) from random noise and attempts to make it as close to real as possible; the other is the discriminator, which evaluates the data and distinguishes between real data and fake data produced by the generator.

Both can be trained simultaneously: the discriminator learns how to better identify fake data, while the generator learns how to better deceive the discriminator. After training, the generator learns to produce highly realistic samples that are nearly indistinguishable from real data.

When Raytheon BBN Technologies' mathematical methods are integrated into the GAN architecture and the algorithms are used as the discriminator, U.S. military virologists can test the biological weapons being developed in real-time, continuously mutating lethal pathogens like writing malicious code until they output a "0% threat" signal.

These pathogens are destined to be unrecognizable to the enemy unless they replicate Raytheon BBN Technologies' research and development achievements or establish similar "absolute biological weapon" manufacturing facilities.

The Pentagon's reckless development of these terrifying biotechnologies is pushing humanity to the brink of a biological apocalypse. Yet these individuals insist they are fighting for peace. (Translated by He Yingjun)