Strengthening biosecurity in the age of advanced AI
Our mission is to advance the scientific and policy breakthroughs needed to reduce catastrophic biological risk, and to accelerate the people and organisations capable of carrying that work forward.
Our ambition is to help build a biosecurity field capable of responding as biological and AI capabilities change rapidly: technically sophisticated enough to understand emerging risks, operational enough to build and deploy practical defences with urgency, and institutionally strong enough to sustain that work as the landscape continues to change. The AIxBio Fellowship is one mechanism for doing this, and it is built around three mutually reinforcing goals:
Advance research
Support work which resolves important uncertainties, tests proposed safeguards and produces evidence which changes decisions.Accelerate talent
Give researchers, engineers, scientists, policy specialists and operators the support and context they need to take sustained responsibility for important biosecurity problems.Incubate organisations
Help promising projects move beyond a fellowship: into existing institutions where they can be deployed, into longer-term research, or into new organisations where an important problem otherwise lacks an owner.
We think this is an unusually important time to be doing this work (read more below on why!). AI capabilities relevant to biology are advancing quickly, while many of the safeguards and institutions needed to keep pace are still maturing. Our AIxBio programmes exists to help close that gap: by backing strong people, important research and new organisations early.
AI Capabilities in the Life Sciences Are Rapidly Expanding
Over the past few years, AI systems have demonstrated remarkable progress in biology and chemistry. Models can now predict protein structures (AlphaFold), design functional genomes (Evo2), and generate new molecular compounds with therapeutic potential, including antibodies (Germinal).
A 2023 paper from The Centre for Long-Term Resilience warned:
“Artificial intelligence and biotechnology are converging in a way that could catalyse immense progress from areas like personalised medicine to sustainable agriculture—as well as substantial risks. There is a potential for new capabilities that threaten national security, including those that may lower barriers to the misuse of biological agents.”
Advances in AI and biotechnology are expanding what individuals and organisations can do with biological systems. Frontier AI models are becoming more capable at biological reasoning, experimental planning and scientific research, as documented in the UK AI Security Institute’s Frontier AI Trends Report and Anthropic’s work on AI and biological risk.
Meanwhile, advances such as AlphaFold 3 illustrate how AI is also expanding what is possible in biological modelling and design. This is rapidly reshaping the biotechnology and biosecurity landscape, enabling unprecedented discovery, but also increasing the risk of deliberate or accidental misuse.
Improving capabilities come with increasing biosecurity risk
Advances in AIxBio are also changing the scale and character of biological risk: lowering barriers to misuse, creating new vulnerabilities, and placing greater demands on systems designed to prevent, detect and respond to biological threats. Frontier AI developers are increasingly treating biological capabilities as a distinct risk domain; OpenAI’s work on preparing for future AI capabilities in biology gives a useful overview of both the scientific opportunities and the dual-use problem.
“Biology is by far the area I’m most worried about, because of its very large potential for destruction and the difficulty of defending against it.”
— Dario Amodei, CEO of Anthropic, 2026
There is substantial uncertainty about exactly how quickly these risks will develop. At ERA, we do not think that uncertainty is a reason to wait. Strengthening biosecurity for the age of advanced AI will take time to build, and the consequences of discovering critical weaknesses too late could be catastrophically severe. This uncertainty, and the case for developing evaluations and safeguards before capabilities cross dangerous thresholds, is discussed explicitly by several frontier labs.
To navigate these emerging complexities, we need a strong field with integrated knowledge spanning artificial intelligence, life sciences, governance and regulatory frameworks. ERA, alongside the Cambridge Biosecurity Hub (CBH), is helping to build the expertise and networks necessary to mitigate catastrophic biosecurity risks in the age of increasingly capable AI and biotechnology.
Defence in depth
Capability & risk assessment
AI safeguards
Synthesis screening
Early detection
Physical resilience
Medical countermeasures
Governance
Biosecurity infrastructure
Biosecurity requires defence-in-depth
There is unlikely to be a single intervention which strengthens biosecurity for the age of advanced AI. Biological risk can emerge at several points: through access to increasingly capable AI systems, the acquisition or synthesis of biological material, work conducted in laboratories, failures to detect an emerging outbreak, or an inability to respond quickly once one begins. A robust biosecurity strategy therefore needs overlapping layers of prevention, detection and response.
“The shift that the world needs to make for AI security generally, and AI biosecurity in particular, is to move from a strategy of blocking to one of resilience.”
— Sam Altman, CEO of OpenAI, 2026
Upstream, this means understanding which emerging AI capabilities actually change biological risk, and building safeguards that make dangerous use harder without unnecessarily constraining legitimate science, for example by combining access controls, with things like real-time classifiers, monitoring and post-hoc detection rather than relying on any single safeguard. As capabilities advance, evaluating whether those systems remain robust under realistic attempts at misuse will become increasingly important.
Biosecurity also has to extend beyond the model itself. Nucleic acid synthesis screening can make it harder to obtain concerning biological sequences, while strong laboratory biosafety and biosecurity can reduce the risk of accidental or deliberate release. The World Health Organization has increasingly emphasised systematic management of these risks, including through its new Biorisk Implementation and Evaluation Tool, designed to help institutions assess and strengthen biorisk management and dual-use research governance.
These interventions operate at different points in the risk pathway and fail in different ways, which is precisely why defence in depth matters. Earlier detection can dramatically change the options available during an outbreak: approaches such as wastewater and environmental surveillance can provide population-level signals before traditional clinical surveillance catches every case. Meanwhile, better respiratory protection can provide pathogen-agnostic protection while more specific countermeasures are developed, and institutions such as BARDA invest in the vaccines, therapeutics, diagnostics and manufacturing capacity needed for rapid medical countermeasure deployment.
The bottleneck is not only ideas but also ownership.
Many important biosecurity problems sit between established fields and institutions. They require biology and machine learning, or science and policy, or technical work and operational deployment. This often means that no single organisation or individual clearly owns them. We think strengthening biosecurity depends heavily on capable people who are willing to take responsibility for these gaps: to understand a problem deeply, identify what would materially change it, and keep pushing until useful work exists in the world.
Sometimes, this means producing research which changes an important belief or decision. Sometimes, it means joining an organisation that already has the ability to act. Sometimes, it means building a new organisation because nobody else owns the problem end-to-end.
Our AIxBio programme exists to help more people do all three.
Interested in learning more?
Selected papers, practical proposals and new perspectives on AI and biosecurity.
Explore the resources