Biosecurity in the Age of AI

As I write this at the end of February 2023, a cascade of articles in the popular press are ringing alarms about the?‘AI Arms Race.’?The risk of major tech companies repeating mistakes made during the launch of social media – unleashing a tidal wave of explosive growth that has trumped long-term public safety – is high. Amidst this steady drum beat of warnings –??including one from Elon Musk?–??ChatGPT, Bard, and other AI platforms are transforming our ways of working, exciting curiosity and experimentation even as we breathlessly attempt to decipher the intrinsic differences between human and machine creation.?

While one can easily imagine (or observe in real-time) the benefits of generative artificial intelligence, the risk space is equally apparent, leaving many of us alarmed by the lack of governance around a sector changing overnight - one in which global tech companies wield the power of nation states.

Even more concerning, this explosion in AI technology is converging with rapid transformation in other sciences, including in synthetic biology. Synthetic biology is a field of science involved with redesigning organisms for useful purposes by engineering them to have new abilities. Concerns about this confluence are also appearing in the?popular press, as reporters weigh the downside consequences of this technology - such as the intentional release of enhanced viruses - against benefits, like faster development of countermeasures such as vaccines and therapeutics.?

Given the catalytic effect of AI on synthetic biology, the time is ‘now’ for rapid action.?As just one example, we have seen the life-saving benefits of CRISPR technology applied to an ever-widening pool of diseases?and treatments, while observing the attendant risks of?tools like CRISPR potentially making viruses more transmissible and/or pathogenic.?

The Biden Administration is taking active steps to develop policy and regulation that enhance both biosafety (preventing accidents) and biosecurity (preventing intentional misuse).?Fortunately,?they are doing so in close partnership with the private sector, whom they recognize as essential drivers of a safe, secure, and thriving bioeconomy - especially in the uncharted waters of biotechnology powered by AI and machine learning.?

Last September, President Biden released an?Executive Order on the Bioeconomy, which called for increased levels of investment into biomanufacturing and biotechnology, as well as improvements to biosafety and biosecurity. The EO outlined an objective to “secure and protect the United States bioeconomy by adopting a forward?looking, proactive approach to assessing and anticipating threats, risks, and potential vulnerabilities.”

The EO goes on to require that this should be by “partnering with the private sector and other relevant stakeholders to jointly mitigate risks to protect technology leadership and economic competitiveness.”

The Biden team seems to understand that technology (when appropriately guardrailed) is the way forward – not something to cut off or inhibit.?

The White House also published?National Security Memorandum 15 on Countering Biological Threats and the much more detailed?National Biodefense Strategy and Implementation Plan. These strategy documents further prioritized biosecurity and called on all parts of government – from the Departments of State to Homeland Security to Defense to the Intelligence Community –??to step up their focus on biosecurity.

This policy momentum is encouraging, but it is not enough.

As it is considering, I would encourage the Biden Administration to bolster these?policy initiatives by advancing draft recommendations recently published by the National Sciences Advisory Board for Biosecurity. While imperfect, these guidelines significantly update and expand definitions for what constitutes ‘risky pathogens’ and end ‘carve outs’ for work on vaccines and surveillance. In addition, they call for similar rules for animal and plant pathogens which are essential given the importance of “OneHealth” approaches and ongoing?research?on animal-human cross over.??

At the same time, there remains a gap in?oversight?of related research for experiments funded by private entities like venture capital firms and pharmaceutical companies. The government should develop restrictions to address this gap in oversight.?

Collectively, we also need to fine tune our risk calculations around synthetic biology – and explicitly communicate those risks to the public.??In 2014, as one example,???Gryphon Scientific??attempted to quantify the risks and benefits of conducting research on Gain of Function?(enhanced pathogens) over wild-type pathogens. They looked at biosafety risks (‘lab leaks’) and biosecurity risks from both accidents and intentional misuse of the information. Though their analysis was considered too broad and perhaps unwieldy by some, Gryphon’s effort was an example of the kind of analysis that could be critical in delivering us from generalities toward thoughtful, middle ground dialogue that enables nuanced responses.?

In addition, HHS and similar bodies should address gaps in transparency – helping us all understand how they’ve made assessments around enhanced pathogen research.?

One key area where a convergence of technologies could help solve, rather than increase, challenges is in so-called “attribution science,” which allows us to determine whether a virus was engineered and if so, to identify which parties were responsible for its creation. The COVID-19 ‘origin controversy’-?which is once again in the news given the Department of Energy’s new classified report- is a great example of what happens in the absence of clear attribution capabilities, but it is not the only one. Russia has alleged that the?U.S. created Monkeypox?and also that?Ukraine is using bioweapons.?While not an attribution issue, the?concern over whether the Chinese spy balloon might have been carrying bioweapons?demonstrates the kind of rampant speculation that can attend the bogeyman of biowarfare in all its forms.

Experts?like Tom Inglesby suggest that we need technical advances in fields like proteomics (protein sciences), glycomics, and epigenetics to improve attribution science. We could also leverage major advances in AI and machine learning to support faster and more accurate sequencing and attribution. In addition, scholars like?Rebecca Katz?have suggested that utilization of ‘blockchain’ type technology might help overcome ‘chain of custody’ challenges around data.

Global leaders should engage with countries progressing bioengineering and convergence technologies, supporting advancements in medical science and agriculture while providing oversight to limit accidents and intentional misuse of these technologies.?

Action should also be undertaken in collaboration with emerging economies to keep dual use technology and enhanced pathogens out of the hands of bad actors, especially in fragile states. (While a?few African and Asian countries have prioritized guidelines for dual use research and enhanced pathogens, many have not.) Thankfully, as the Biden team works with global leaders to develop their economies while also helping implement safeguards, such engagement already seems to be part of their approach.?

All of these measures will be inadequate, however, unless national and global governments can resolve ‘chain of command’ and oversight challenges that have persisted in the field of biosecurity for decades.

Indeed, effective biosecurity responses could be jeopardized by the multitude of organizations responsible for investigation and response. The “UN Secretary General’s Mechanism”??– (UNSGM) is the UN entity charged with investigating uses of biological weapons and relies on a network of?labs and consultants?to sequence pathogens and determine origin and responsibility. It operates under the umbrella of the UN Office for Disarmament Affairs (UNODA).?

The duties of the WHO’s Health Emergency Programmes office, meanwhile, also include investigations into the causes and origins of pandemics and health emergencies. We will need to work collectively to ensure these organizations are ready to engage smoothly and efficiently, especially during biosecurity emergencies. Without clearer governance and specified ‘chains of command’ in the U.S. and globally, significant time could be lost, credibility undermined, and mistakes made during a biological crisis.?

Finally, all of these efforts need to align with the ongoing work by the Intergovernmental Negotiating Body (INB) to draft an accord on pandemic prevention. At present, the ‘Zero Draft’ calls on countries to strengthen oversight and reporting on labs conducting dual-use or gain of function research, but without suggesting any compliance mechanism.?

The world is woefully behind in governing AI. Its convergence with biosecurity should create a call to action for lawmakers and global leaders. In the U.S., we need a bipartisan, multi-sector coalition to evaluate this convergence and take dramatic steps forward to advance technology and policy. On the global stage, we need much greater coordination between the UN, the WHO, and regional bodies like the African CDC. And we need this coordination to be done in the context of the pandemic accord and the ability to outline a better path forward.?

Such action to enhance coordination at this critical juncture could pave the way for more robust AI and biosecurity regulation overall. It could prevent the most devastating effects of exponential technology - technology currently left unchecked by those with the power to safeguard it.?