Image source: SEFH
Biosecurity has an AI problem. And AI has a biosecurity problem. According to a few scientists, engineers and AI specialists, the use of artificial intelligence in life sciences is a double-edged sword. While AI promises great potential for biotechnology solutions in medicine, agriculture and sustainability, the new capabilities created by AI also pose a risk of misuse. Determining how to manage the dual-use nature of AI is a growing security issue.
Although the current impact of AI on biological threats is minimal, regulatory gaps are concerning. Implementing legislation requiring – rather than simply recommending – oversight and red teaming in AI development would help close these gaps. Strengthening regulation of manufacturing supply chains, specifically designed to address the intersection of AI and biotechnology, would also help rebuild barriers to the physical development of biological weapons. Ultimately, the impact of AI on the evolution of life sciences requires a reassessment of the biological threat landscape and the governance needed to address it.
New technology, old threats?
In March 2023, Vice President Kamala Harris radius on the future of AI, recognizing that it “has the potential to do profound good.” She also referred to the threat posed by AI-formulated biological weapons to “the existence of humanity.” The tech industry agrees: Eric Schmidt, former Google CEO. called The most pressing problem with AI is its potential use in biological warfare.
Biological weapons are micro-organisms – viruses, bacteria and fungi – or toxins produced by living organisms that are “deliberately released to cause disease and death in humans, animals or plants”. Of course, biological warfare is not new. Several countries experimented with biological weapons programs during both world wars. The Soviet Union stockpiled biological weapons throughout the Cold War. Anthrax attacks killed five people in the United States in 2001. Research shows that the current state of AI will not necessarily exacerbate existing capabilities. Nonetheless, the dual-use nature of AI raises concerns that tools intended for biomedical research could be repurposed in harmful ways and open the door to biological threats that did not exist before.
Dual-use AI and biosecurity
Large Language Models (LLM) and Biological Design Tools (BDT) are AI systems being scrutinized, among other things, for their prospective use in the creation of biological weapons. LLMsuch as OpenAI’s Chat GPT-4, are trained on huge amounts of text data to learn how to generate and process text in a process called deep learning. While LLMs are trained in natural language, BDT are trained on biological data. BDT can “design new proteins or other biological agents”, helping users to better understand and manipulate biological systems.
These AI tools alone cannot create biological weapons, but can simplify the process by providing more easily accessible information. This democratization of knowledge lowers the barriers to entry into the field and characterizes the double-edged nature of AI in biosecurity. LLMs, for example, help biological and chemical research and reduce the time and cost of drug discovery. But they could also benefit novices by consolidating online sources on biological warfare and refining the results into easy-to-digest terms. BDT significantly advantage life sciences and can accelerate progress in areas such as vaccine development. However, their application is also subject to abuse: the same technology used to suggest chemicals for medical purposes can be used to produce chemicals. harmful substances like VX, one of the most powerful poisons in the world. Ultimately, political leaders fear that state or non-state actors interested in developing unconventional weapons capabilities will take advantage of this dual-use technology.
Governance and regulatory challenges
AI is difficult govern – this is progressing rapidly and crossing industries and borders. There is no comprehensive federal legislation regulating AI in the United States, and governance is often vague and overarching, relying on self-regulation by companies and institutions. The Biden administration has issued extensive decree in October 2023, outlining standards for AI developers. In biotechnology, this included recommendations to meet reporting requirements and safety testing. Since the decree, many companies have implemented AI red-teaming – an organized attack on systems to test and improve security – and funds research into how to make their systems more secure.
In a agreement signed by more than ninety scientists and biologists in the field of AI, the parties agreed to take precautionary measures against any misuse of their work. Anthropic, leading AI company work with biosecurity experts throughout the development of its chatbot, Claude. OpenAI takes steps to prevent misuse of its chatbots. The organization released an open letter in January 2024 providing the public with information about the red team. research conducted in response to the fact that its latest model, GPT-4, offers a “slight improvement in the accuracy of creating biological threats.”
However, some AI developers believe they are only partially responsible for mitigating biosecurity threats. David Baker, director of the Institute for Protein Design at the University of Washington, think the “proper place to regulate” lies in the physical barriers to biological weapons: the laboratories and equipment that enable processes like DNA synthesis. DNA synthesiss has been used in various applications, including vaccine development, but can also provide components against harmful pathogens. Biden’s executive order recommends that suppliers of DNA synthesis equipment audit their purchases to monitor abuse; however, there is no real requirements that suppliers must comply with this recommendation.
Implementing legislation that mandates appropriate oversight throughout supply chains will strengthen biosecurity and close gaps in AI regulation. For example, European AI law governs AI technology based on its risk level classification. However, with some dual-use AI models, it is difficult to anticipate and manage risks. LLMs are difficult to predict: their abilities evolve so quickly that new and dangerous uses are often not detected until after the LLM has been used. On the other hand, excessive regulation risks stifling the benefits of AI biotechnology. Thus, policy to regulate specific areas of physical production of biological weapons can help address AI governance issues. DNA synthesis, for example, requires materials that are inexpensive and accessible. Targeting industries known to be more accessible can rebuild the barriers to creating biological weapons.
Prepare for the future
We are at an inflection point for AI and biosecurity in a rapidly evolving security environment. America’s meager biodefense capabilities leave many not convinced of its ability to counter large-scale attacks, especially as the response to the COVID-19 pandemic has heightened fears that a significant biological attack could lead to widespread chaos and human suffering. On the bright side, experts suggest that we are not there yet. Even if the possibility for AI to influence a biological disaster exists, its current impact on biological threats is minimal.
Experts are rightly allaying fears about large-scale attacks by biological weapons while taking very seriously the threat that AI-based technology could pose. To address potential risks, more concrete legislation emphasizing red teaming and compliance with standards and regulations will help demystify the convergence of biosecurity and AI. Strengthening regulations on laboratories, equipment, and other resources needed to create and stockpile harmful biological weapons, particularly those related to AI capabilities, could also help prevent biological threats. Preserving the beneficial use of AI in the life sciences will require a delicate balance. Managing the double-edged potential of AI in biotechnology will be a long and complex task, with global ramifications, but one worth monitoring as AI advances.
The opinions expressed are those of the author.