Is AI creating bioweapons, or are we making them with AI?


Advances in technologies such as AI and quantum computing are beginning to create opportunities to change our mutual home for the better. But on the flip side, stories of AI-made bioweapons are becoming too close to reality.

AI brings opportunities for energy grid optimization and hastening clean energy development in the fight against climate change. It can also be used in agriculture for crop yield optimization and to make data-based decisions that assist farmers in maximizing food production for an ever-increasingly hungry population.

While so many are dedicated to preserving our planet, some people seem intent on destroying it. Be it through the autonomous weapons system picking out our demise with killer robots, AI-controlled drones used within military forces, or robot soldiers on the battlefield, our worst adversary seems to be ourselves. The question is, how have we reached this point?

ADVERTISEMENT

From Unit 731 to AI: how nations weaponized disease

Unfortunately, there’s a long history of viruses or bacteria being used as weapons in war, and very few regions can claim the moral high ground. Japan's infamous Unit 731 weaponized typhus and cholera in World War II, bringing chemical warfare into the battle. Elsewhere in the former Soviet Union during the Cold War, a bioweapon program involved the manufacturing of smallpox, plague, anthrax, and even drug-resistant plague.

Between 1949 and 1969, the US military also developed biological weapons that weaponized biological agents such as anthrax and botulinum toxin. President Richard Nixon ended the weapons program in 1969, and 55 years later, technological advances are returning governments down the dark pathways of the past rather than forward as a progressive species.

How AI is lowering the bar for mass destruction

Biochemist Rocco Casagrande recently demonstrated to White House officials how AI chatbots like Gemini, Claude, and ChatGPT could assist in developing dangerous pathogens. He warned that the same tools that help users find recipes that match the ingredients in their homes could enable the creation of bioweapons in basements and high school labs.

The White House has already secured voluntary commitments from all the big AI companies to assess and manage AI risks in biosecurity. This move paved the way for last year's global AI Safety Summit in London, where Vice President Kamala Harris warned that AI-enhanced bioweapons could endanger humanity's existence.

The dangerous ease of designing bioweapons with AI

These concerns were also supported by a recent study highlighting the alarming potential for AI models, initially designed for drug discovery, to be repurposed for bioweapons. Within six hours, an AI was able to generate 40,000 potentially toxic molecules, including known chemical warfare agents and new compounds predicted to be even more lethal.

ADVERTISEMENT

The rapid and unintended capability raises severe concerns about the ease with which such AI tools can be misused. The study also demonstrates that designing toxic molecules requires minimal time, effort, and resources, underscoring such technology's efficiency and danger.

The authors emphasize the need for more oversight in the availability of AI tools and datasets, which could lower the barriers to creating biochemical weapons. Many also call for ethical training in AI, especially for students, to raise awareness about these risks.

The study's authors recommend fostering dialogue at scientific conferences, implementing a code of conduct for AI companies, and enhancing security measures for public-facing APIs to protect humanity from itself. However, the most significant challenges ahead are more complicated than those caused by a psychotic loner in a basement with AI tools.

AGI 2029 and Terminator 40 years later

Forty years have passed since the Terminator movie warned that AI would enable machines to become more intelligent than humans and begin to take over the world in 2029. Ironically, in 2024, our reality sees respected leaders from Nvidia CEO Jensen Huang to renowned futurist Ray Kurzweil predicting we will reach artificial general intelligence (AGI) by 2029.

Rather than AI being the villain, a better question might be what makes us so surprised by our reflection. Biological warfare is as old as 1763-64, when the British gave smallpox-contaminated blankets to Shawnee and Lenape (Delaware) communities.

Is AI the digital boogeyman or a mirror of humanity?

The problem is not technological prowess but how it exaggerates humanity's worst traits. Being an eternal optimist, I believe there are more good than bad people on this earth, but there is no hiding from the fact that language-based AI learns from the information created by humans.

So, if we see AI promoting war, cruelty, intolerance, hatred, greed, and destruction, we are merely being presented with a reflection of ourselves. After all, we are all an anagram of our parents.

ADVERTISEMENT

Over 100 scientists have committed to policies to avoid using AI in bioweapons to ease concerns about individuals or malicious groups creating bioweapons from basements. However, history would suggest that nation-states are more likely to break the rules and weaponize biology.

Rather than being accountable or taking responsibility, many on the wrong path will find it easier to blame AI or any new technology. But it's humans and not AI that will create the next pandemic or threaten to unleash the biological weapons that could threaten our existence.

How we break this cycle of self-destruction and change the world for the better is much more of a humanity problem than AI being a digital boogeyman.