Background

Oxford Biosecurity Group (OBG) conducts global research projects in collaboration with technical and policy organisations, to foster talent and tackle issues to reduce global catastrophic biological risks (GCBRs) and risks at the intersection of AI and biology.  

Each project is led by a project lead, with 3-6 researchers each spending 5-10 hours a week doing research. Past projects include AI-bio, synthetic DNA regulation policy, and lab safety governance, with collaborators such as 1Day Sooner, CEPI, and Center for AI Safety. Researchers and project leads can gain skills, experience and networks in biosecurity, and promising researchers are referred by us for additional support and opportunities. Projects are typically run in ‘project cycles’ of 5 or more projects, which allows for economies of scale for operational and time costs for processes such as recruitment and onboarding, and for easier resource sharing and networking across projects. Since our founding in October 2023, we have run 3 project cycles including our pilot cycle. In the past, projects have been 7-8 weeks long, but are planned to be extended to 10-12 weeks for future project cycles. 

Past Impact

In the two project cycles after our pilot, 51 researchers completed 14 projects run by 12 project leads with 8 collaborating organisations or initiatives. We identified 31 promising participants for further support and opportunities. The next steps of past participants include Bluedot Impact (facilitator, participant and teaching fellow), Existential Risk Laboratory, Oxford Nanopore Technologies, relevant study, founding and directing a non-profit, and continuing with full-time or part-time work with the collaborating organisation. Project outputs have been presented at the International Pandemic Sciences Conference 2024, fed into collaborator strategies, and three projects are planned to feed into academic publications. For example, the results of the project ‘Eliciting Biological Knowledge of AI Models’ were uploaded as a preprint, which has received significant interest from relevant stakeholders and was used by OpenAI in their o1 system card biosecurity threat evaluation (see section 4.5.7 of the system card here).

For the full list of our past project outcomes see here. Our full list of participant testimonials and next steps is not shareable for privacy reasons, but we can give some further details privately upon request. We will release a more detailed impact evaluation in the next few weeks. 

We expect most of our impact to come from identifying promising people, who might not otherwise be identified (e.g. people who are more interested in projects than courses, and those in their mid-career looking for object-level work); and helping them gain skills, experience and connections. We also enable previously identified promising people to gain experience doing biosecurity work, which can help them when applying for future jobs and grants. Direct project impact is hits based and is likely to come from increasing the capacity of relevant organisations, which allows additional valuable work to be done, or to be done sooner. 

Early 2025 Plans and Use of Funding

We plan to run our next main project cycle starting in early 2025, and for this, we will initially run a project lead recruitment round. The aim of this is to attract entrepreneurial, operations and managerial people, which have been identified as a skill gap within biosecurity. We are also scoping out additional projects, including projects specifically targeting more experienced professionals, which we will announce soon.  

To confirm the dates of the project cycle, we are fundraising for a minimum of £13,300/$16,700, which will cover the remaining costs of a bootstrapped version. This is primarily to cover project lead stipends and project-specific costs, e.g. compute for AI-bio projects, and will be used to run additional projects if the full amount is received from outstanding funding applications. 

Thank you very much for your support, and happy to answer your questions via comment or direct message.