AIxBio Winter ’27 Tracks
Work on a concrete biosecurity problem through one of three routes at ERA this winter: investigate an open research question, embed in an organisational team already working on it, or test whether a neglected gap warrants a new organisation. Across all three of our tracks, our main goal is to help you make meaningful progress on mitigating catastrophic biological risks, including those amplified by advances in AI and biotechnology.
Research an open question that matters for biosecurity.
Work with an expert mentor on a sharply scoped scientific, technical or policy question relevant to reducing biological risk. You might extend an existing research agenda, test an assumption that has not been examined closely enough, or build evidence in an area where important decisions are still being made under substantial uncertainty.
The bar is whether the work changes an important belief or decision. Strong projects should leave behind something other researchers, labs or policymakers can actually use: a better evaluation, an empirical result, a dataset, a technical prototype, a threat-model analysis or decision-relevant evidence.
Explore the research landscapeIllustrative project directions
-
Biological capability evaluations
Where do current biological capability evaluations succeed or fail? Test whether benchmark performance tracks meaningful increases in real-world capability, and develop evaluation methods that better capture AI-enabled biological risk.
-
Safeguards for biological AI
Assess safeguards around biological models and design tools. Identify failure modes, test how protections behave under realistic conditions, and generate evidence about which safeguards meaningfully reduce risk.
-
Evidence for governance
What technical evidence should inform decisions about model access, deployment or release? Develop analyses that connect biological risk measurements to workable standards, thresholds or oversight mechanisms.
Contribute to a real workstream inside a biosecurity team.
Join an organisation or research team and take responsibility for a bounded piece of work on an ongoing project. In this case, projects will be scoped against live priorities, with supervision from staff across our placement organisations.
A placement may be scientific, technical, policy-focused or operational. Unlike a standalone research project, success is defined by usefulness to the team: shipping an analysis, building evaluation infrastructure, improving a screening workflow, or resolving a decision the organisation is currently blocked on.
Fellows normally work from ERA’s Cambridge office. The exact level of embedding, supervision and any time spent with a host organisation are agreed case by case.
Illustrative contributions within a team
-
Evaluation infrastructure
Build the tooling, pipelines or analysis workflows an evaluation team needs to turn biological model testing into reliable, decision-relevant evidence.
-
Synthesis screening
Contribute to nucleic acid sequence screening, customer verification or screening implementation — for example by improving tooling, testing coverage or solving a concrete deployment bottleneck.
-
Technical and policy strategy
Work through a decision where an organisation needs stronger evidence: whether to prioritise a new research area, adopt a safeguard, support a standard or invest in a larger programme of work.
Test whether a neglected biosecurity gap deserves an organisation.
Some important problems persist because no existing team clearly owns them. The Founding track is for testing whether one of those gaps merits a dedicated organisation or programme — and whether you are well placed to build it.
Start with the problem, not the institution. Map existing efforts, speak to the people who would use or support the work, pressure-test the theory of change, and build a small prototype or pilot where useful. By the end of the fellowship, you should have a much clearer view of what needs to exist, who needs it, what capabilities it would require, and whether the idea should continue.
A good outcome is sometimes deciding not to found something. We are excited to help you establish whether a new effort would add real biosecurity value, and if yes, we can provide strong incubation support.
Illustrative directions for new initiatives
-
Independent safeguard assessment
Could independent assessment of biological AI safeguards improve decisions by labs, policymakers or other actors? Test who would use the work, what evidence they need, and what would make an assessment credible.
-
Screening implementation
Identify a concrete bottleneck preventing effective synthesis screening from being adopted or implemented well, then test a tool, service or intervention with the organisations facing it.
-
Bioforensics and attribution
Map an important gap in biological forensics or attribution — whether in methods, data, capacity or institutional ownership — and define an initial research programme that could materially improve investigation of unusual biological events.
Support across every track
Our programme is built around concrete AIxBiosecurity projects. We will be scoping directions with mentors and teams, then recruiting fellows whose expertise can either materially advance existing work or bring new perspectives to the field. Both the projects and mentor / placement matches will be refined together with your Research Manager before the fellowship begins.
Expert guidance
Mentors can help sharpen research questions, catch bad assumptions early, and connect the work to the wider scientific, technical and policy context in the field.
Dedicated research management
Our Research Managers are themselves experts in the field. They can help keep projects moving, work through bottlenecks with fellows, and broadly support the fellows throughout the ten weeks & beyond.
Time and practical resources
Fellow funding, Cambridge workspace, operational support and access to compute for projects that need it.
Working arrangements
The right working arrangement depends on the problem, the team and where the output will be most useful.
Can a placement involve research?
Yes. Many placements will be research-heavy. The distinction is organisational: placement fellows are embedded in an existing team’s agenda and produce work for that team, while Research track fellows primarily pursue a standalone mentor-guided question.
Do outputs need to be public?
No. Some biosecurity work is more useful, or more responsible, when it remains internal due to infohazard considerations. Projects can both produce public research or private analysis for an organisation’s decisions.
What happens after the ten weeks?
We are excited to support top fellows with up to six months of further support through ERA’s Research Scholar programme, allowing promising research, collaborations or projects to continue beyond the fellowship.