By Patrick Norén
From October 22-24, the Organisation for the Prohibition of Chemical Weapons and the Kingdom of Morocco organized the first ever conference on the role of artificial intelligence (AI) in advancing the implementation of the Chemical Weapons Convention.
This landmark event, which I attended in my capacity as editor of CBNW Magazine, welcomed over 200 attendees from 46 States Parties to the Chemical Weapons Convention, including representatives of national and international governmental organizations, national authorities, policymakers, scientific experts, industry leaders, and journalists.
Spread over three days, the conference covered a wide range of topics, including those directly related to AI’s role in advancing the Convention’s implementation, as well as broader issues surrounding the evolution and increasing use of AI in diverse fields beyond the OPCW’s mandate.
A Catalyst
Numerous speakers throughout the event remarked that the likes of generative AI and large language models are a “catalyst”. Indeed, if one takes the scientific definition of a catalyst as being a “substance that enables a chemical reaction to proceed at a usually faster rate or under different conditions than otherwise possible”, AI is and will continue to be its most powerful when used in concert with existing processes and technologies to optimize them.
The pedant may argue that AI is not a catalyst, however. After all, catalysts speed up chemical reactions without undergoing any substantial change themselves. AI on the other hand is changing, evolving, and advancing at an astonishing rate, with its computational speed roughly doubling once every three to four months. Putting scientific specificities aside, however, what cannot be disputed is AI’s ability to revolutionize almost every facet of human existence.
AI as a Dual-Use Technology
While AI’s power and potential are already appreciated by the early adopters of the likes of ChatGPT who might turn to it for help with writing essays, computer coding, looking up recipes, date ideas, and so on, what the conference in Morocco impressed upon me most was the truly astonishing breadth of AI’s more sophisticated applications.
AI is already being used in industries related to the implementation of the Convention. For example, AI is already being used to track the sale and shipment of hazardous chemicals, to identify suspicious transactions, and to identify and synthesize new chemical compounds to be used in medicinal drugs. Although unrelated to the Convention, conference attendees also learned about AI’s current application in agriculture, including in predictive analysis of weather and resource allocation.
However, a theme throughout the conference was AI’s status as a dual-use, enabling technology. That is to say, for each way that AI could be harnessed for the good of humanity, there exists another way that AI could be harnessed for the opposite. For example, those same technologies that are already being used to identify new chemical compounds for all sorts of industries could also be used by nefarious actors to develop new, cheaper, and more potent chemical weapons. The same AI technologies that are designed to tighten control and verification regimes could also be used to find ways to circumvent those same regimes.
It was noted in the conference that humans have already put some safeguards in place in certain AI programs to try to prevent malign actors from using them to access and utilize information for nefarious purposes. However, these safeguards can often be circumvented with a little imagination, and the extent to which AI programs permit access to potentially weaponizable information remains under human control.
Legislation, Regulation, Cooperation
Science does not exist in a vacuum, however. It was for this reason that the conference united both scientists and policymakers to start the discussion on how to regulate AI so that it advances human security as much as possible while not endangering it by equal measure.
It was noted at the conference that while science had responded very quickly to COVID-19 – with the earliest vaccines being administered only a matter of months after the global outbreak of the virus – policymakers have not. Indeed, legislation and regulation on artificial intelligence is still lacking, bar a couple of notable exceptions such as the European Union’s AI Act.
But making policy to govern and regulate AI comes with quite unique challenges. One is the problem of concretely defining what AI is and what it does. It is a relatively new technology that is evolving at a rapid rate, and it is likely that AI’s evolution will continuously rapidly outpace humanity’s ability to regulate it. For example, laws on AI painstakingly written over months or even years could be rendered outdated or redundant after no time at all, necessitating the rewriting of those laws.
If humanity’s regulation of AI is unable to keep up to pace with its evolution, could humanity, I asked myself, use AI to help regulate AI? Could AI drastically reduce the time it takes to create standards and regulations governing itself? What are the potential ethical implications of this?
Another challenge is the issue of enforcement. If, hypothetically, a Chemical Weapons Convention-style treaty were designed to govern global AI development and employment, how does one convince the politicians to potentially surrender control over a technology that could be used to the great benefit of their population? Or, perhaps, to the great detriment of their enemies? With what serious teeth would any implementing organization be endowed?
It is worth highlighting that the issue of enforcement is something with which the OPCW struggles. 2023 saw the destruction of the final declared stockpile of chemical weapons by the United States of America, and yet allegations of chemical weapons use by Russia in Ukraine continue unabated. As was noted by attendees at the conference, self-declarations, transparency, and mutual trust between States Parties and the OPCW are central to the organization’s ability to effectively implement the Chemical Weapons Convention.
Misinformation and Disinformation
I am neither a chemical scientist, nor am I a seasoned diplomat. Rather, my experience and expertise are in multi-disciplinary research, writing, and editorial. Information conveyance is my trade, and for this reason I was pleased to see some attendees at the conference mention the role that AI could play in CBRN misinformation and disinformation. This is especially relevant if a targeted or rapid disinformation campaign aided by AI should precede or follow an attack that contravenes the Chemical Weapons Convention.
However, while the conference was highly successful in uniting the 1% of experts who are outstanding in their field and for whom AI and the Chemical Weapons Convention are part of their daily lives, the conference did not discuss how to address the remaining 99% of the global population for whom highly sophisticated AI and the Chemical Weapons Convention are alien concepts, but are no less important for their safety and security.
As I have written previously, global CBRN education is severely lacking, as evidenced by COVID-19. In the hands of a malign actor, AI has an extraordinary ability to weaponize people’s ignorance of CBRN risks through disinformation that would undermine emergency response in the short term, and erode trust in authorities in the long term.
At a basic level, more attention must be given to how to educate the 99% on CBRN issues, thus reducing the knowledge vacuum that AI-generated CBRN disinformation would seek to fill. At a more advanced level, policymakers should consider how to actively engage the 99% in developments surrounding CBRN, and the very positive role that AI could play in improving CBRN safety and security.
It is equally important to make the 99% aware of the power of CBRN disinformation, and why precisely this type of disinformation may be of interest to a malign actor. The role that AI could play in producing and spreading it, as part of a coordinated attack that contravenes the Chemical Weapons Convention, must also be included in these educational efforts.
One must not lose sight of the fact that AI and the Chemical Weapons Convention were ultimately both conceived for the benefit of humanity, and yet most of humanity has little knowledge of the key issues. Communicating the insightful, interesting, and valuable information shared in the conference with the 99% in an accessible way that complements and encourages adherence to the Chemical Weapons Convention is essential and will go a long way towards preventatively combating potential CBRN disinformation operations in the future.
Can, Could, Might
The above notwithstanding, it was important to recognize that attendees at the conference were often using verbs such as “can”, “could”, and “might”, and not “has been”, “is”, and “will be”.
Although AI is already being employed in many different industries, its use in the fields of chemical security, chemical terrorism, and chemical weapons is still somewhat unchartered territory. Current discussions in this field are largely – although not exclusively – taking place in the realm of what is possible and not what is actual.
Nevertheless, the OPCW’s conference on AI and the implementation of the Chemical Weapons Convention was a well-timed and much-needed initiation of targeted discussions between diplomats, scientists, and others working in a niche but highly consequential field. With a rapidly evolving and disruptive technology such as AI, it is easy to fall behind.
Science has demonstrated, is demonstrating, and will doubtless continue to demonstrate the astounding potential of AI across multiple disciplines. Scientists should be empowered to continue in the same vein while remaining cognizant of AI’s dual-use vulnerabilities.
Meanwhile, governments and governmental organizations who have the power to draft, pass, and enforce policy on AI should be wary not to fall into an endless cycle of discussions devoid of actions. “Can”, “could”, and “might” will one day turn into “has been”, “is”, and “will be”, and it is incumbent on policymakers to ensure that policy is enacted and enforced proactively and not retrospectively.
As I mentioned previously, this is extremely challenging to do because AI is evolving at a speed that far outpaces policymakers’ ability to regulate it. Despite this and other challenges, inaction is not an option, and neither is under-action. At this early stage, there must be an understanding that policymakers’ desire to ensure that AI is used for the benefit of humanity must outpace malign actors’ desire to achieve the opposite.
I would like to thank the OPCW and the Kingdom of Morocco for organizing the Global Conference on the Role of AI in Advancing the Implementation of the Chemical Weapons Convention, and for kindly facilitating my participation in this event.
Patrick Norén is the Editor of CBNW Magazine.
