AI Drug Discovery Systems Might Be Repurposed to Make Chemical Weapons

Recently, Scientific American reported that Collaborations Pharmaceuticals, a private company in North Carolina, put its AI-based MegaSyn software to work generating a compendium of toxic molecules instead of using animal toxicology data. Their research results demonstrated that the AI software generated 40,000 substances, including not only the target molecule but also other molecules for known chemical weapons, as well as many completely new potentially toxic substances. Many AI drug discovery systems used artificial neural networks. The AI systems scored a molecule based on certain criteria, and a higher score represented that the substance might be more likely to have the desired effect.

Collaborations’ MegaSyn was typically used to develop drugs for treating Pitt-Hopkins syndrome or Chagas disease. In this study, the company’s scoring method revealed that many of the novel molecules MegaSyn generated were predicted to be more toxic than VX, a nerve agent. Some experts on biodefense programs commented that the development of actual weapons in past weapons programs had shown time and again that what seemed possible theoretically might not be possible in practice. However, the ease with which an AI could rapidly generate a vast quantity of potentially hazardous substances could still speed up the process of creating lethal bioweapons.

The researchers of this project are afraid if they had crossed an ethical boundary and decided not to do anything further to computationally narrow down the results or not to do any experiments to test these materials. They warned that people should pay more attention to the issue of the AI technology, which might be sidetracked for more nefarious purposes. In addition, they proposed several ways to monitor and control the access of those who could use the AI drug design software to models, data and code.