Artificial Intelligence (AI) has transformed the ways knowledge is generated and experiments are conducted across disciplines. Chemistry is one of the scientific disciplines which is a beneficiary of AI applications, with profound and transformative impact. However, the application of AI in chemistry also poses a dilemma where it can serve both civilian purposes and military ends. On the one hand, it provides opportunities to accelerate research and discoveries in chemistry as well as assist in meeting disarmament goals by enhancing compliance with the Chemical Weapons Convention (CWC), as highlighted in the 2025 AI report by the Organisation for the Prohibition of Chemical Weapons (OPCW) Scientific Advisory Board. On the other hand, it exacerbates existing threats and creates unprecedented risks to the CWC by enabling the development of new lethal and toxic chemicals beyond the scope of the CWC.
The dual-use character of AI applications in chemistry poses risks to current governance mechanisms. The CWC is built on the premise that controlling toxins, production facilities, and precursor chemicals is sufficient to prevent misuse of chemicals. However, AI is shifting the risk from physical to non-physical capabilities, including algorithms, datasets, and computational models. In the “Berlin Conference on the Role of AI in Advancing the Implementation of the CWC” held in June 2024, OPCW Director-General Fernando Arias urged that “the situation today is quite different from when States Parties designed the CWC more than 30 years ago. The implementation of the Convention must be adapted to address the spectacular scientific and technological developments, including Artificial Intelligence.” AI can be employed to design, synthesise and deliver chemical weapons far faster than OPCW can regulate.
AI-enabled Design
Traditionally, formulating new molecules required a team of skilled chemical experts. However, advancements in AI-driven molecular design models now mean that even those with limited expertise can design toxic chemicals without the help of experts. Generative AI models can now access and analyse all available chemical data and generate results that previously required the expertise of PhD-level chemists. Researchers at Collaborations Pharmaceuticals have shown that generative AI and molecular design systems, originally built to accelerate drug discovery, can be used for nefarious ends. The “MegaSyn” system, drug-development AI, was able to generate more than 40,000 toxic molecules in six hours, including novel compounds that could evade existing control mechanisms. AI-powered tools for molecular designing can therefore be used for the research and development of novel chemical weapons if applied unethically.
AI-enabled Synthesis
AI can also assist in synthesising toxic and non-toxic chemicals that can be used as precursors for chemical weapons. Currently, international control mechanisms have effectively established a precursor barrier for non-state actors by strictly regulating not only toxic chemicals but also their precursors. AI-driven retrosynthesis programmes can backtrack a given chemical into its building blocks, suggesting the roadmap and alternate pathways bypassing current control mechanisms. This is crucial because many traditional chemical agents (e.g. nerve agents) were hard to make, precisely because their precursors were controlled. However, as AI can find alternate precursors and synthetic pathways that may use ordinary lab reagents, it can render such mechanisms insufficient to counter the production of chemical weapons.
AI-Driven Delivery Systems
Beyond designing and synthesising toxins, AI-driven systems can also help deliver chemical weapons. The chemical agents now could be dispersed with far greater precision and anonymity than in the past through AI-powered drones. Drones lower the logistical bar for non-state actors to carry out chemical weapons attacks. In 1995, Aum Shinrikyo, a Japanese terrorist organisation, carried out Sarin gas attacks on the Tokyo subway. They had to rely on crude methods; five bags were filled with liquid nerve agents and left in trains, leaving 13 dead and 5800 injured. Now, a remote-control drone equipped with precision sensors could make chemical agent delivery far easier.
Opportunities of AI in Chemistry
AI also has an overwhelmingly positive impact on the field of chemistry when employed for ethical and peaceful purposes. Where AI can be used to produce toxic chemicals, it can also be used to produce more effective antidotes by analysing vast data of molecules for potential antidotes or designing new treatments against novel toxic chemicals. Pharmaceutical companies already use deep learning models to design new, more effective drugs faster, and machine learning models assist scientists in predicting compound toxicity. AI can optimise chemical synthetic processes that produce toxic and hazardous waste and ensure safe disposal of the same.
CWC Legal Framework
AI presents both challenges and opportunities for CWC enforcement. Article I of the CWC explicitly bans each State Party from “developing, producing, acquiring, stockpiling or transferring chemical weapons under any circumstances.” In principle, any use of AI to pursue these prohibited goals would fall under Article I. The CWC regime relies on States Parties declaring their chemical activities and on inspections of potential and suspicious facilities. AI-driven molecular design work is free from any such constraint and can be performed, without the labs, on a computer, anywhere, even by non-state actors.
AI offers various avenues where its application may benefit the CWC enforcement. Machine learning can be used to power sensors that detect chemical weapon threats. AI can automate analysis of chemical samples and use open-source data to identify anomalies in real time. National Authorities can use AI to tighten domestic control over CWC Schedule I and II chemicals (a comprehensive list of toxic chemicals and their precursors, which could be used as or converted into chemical weapons), for example, using AI to analyse industrial production data to detect abnormal patterns and verify declarations by the industries.
Challenges of AI for CWC Enforcement
However, AI-driven molecular design, synthesis, and delivery of toxic chemicals also pose complex challenges for the CWC regime. The pace at which AI drives these processes can overwhelm existing control lists. When a Novichok chemical was used in 2018 in the poisoning of Sergei Skripal, a former Russian intelligence officer who became a double agent for the UK, it took almost 17 months to list Novichok in Scheduled Chemicals subject to declaration and verification under the CWC. AI can develop such toxins at a faster pace that may overburden the CWC procedural framework, inspection and verification regime.
International Response
The international community has already begun to grapple with the potential consequences of the application of AI in the military field, especially weapons of mass destruction (WMD). Between 2024 and 2025, States Parties to the Convention held several high-profile events to address the issue. The “Global Conference on AI and the CWC Implementation,” held in Rabat, Morocco, brought together 200 participants from 46 states to share their research and policy ideas. While underscoring the transformative potential of peaceful uses of AI in the field of chemistry, the Conference notably also cautioned on the misuse of AI by non-state actors.
The OPCW is encouraging the constructive use of AI for verification purposes. Its AI Research Challenge (2025) called for proposals from scientists to strengthen the OPCW’s verification and response capabilities. The OPCW’s Scientific Advisory Board has taken steps to study the implications of AI for the CWC regime, including the establishment of a temporary Working Group on AI.
These developments indicate that the OPCW recognises AI as a priority issue. The CWC should evolve from a mechanism focused on controlling chemicals to one that also governs AI-driven chemical design and synthesis tools. States parties to CWC should contribute to this endeavour by promoting safe AI innovations, such as drug discovery, while remaining alert to threats posed by the dual-use of AI.
