During the project, the researchers will work on the following: 1. A risk assessment of the biosecurity risks of CHIM studies and the infrastructure necessary to conduct them. This will be done through methods such as reviewing the literature and speaking to experts, and will look both at risks that have already been identified and potential additional risks. 2. An evaluation of how these risks apply to specific pathogens and types of pathogens (e.g. advances in pathogen engineering may be higher risk for viruses than bacteria and for certain families of viruses over others). This should include a discussion of pathogen characteristics that increase risk. Researchers should focus first on respiratory pathogens with pandemic potential, such as influenza, SARS-CoV-2, and MERS-CoV, but, where relevant, should include considerations for other diseases such as those on CEPI’s priority pathogen list.), WHO’s priority pathogens list, and Disease X. 3. An evaluation of the relationship between the benefits and risks of CHIMs (briefly discussed in this Pandemic ethics accelerator report, but not deeply assessed yet). 4. A recommendation of methods of mitigating these risks while retaining, as much as possible, the benefits of CHIM studies. 5. A discussion of how this risk/benefit analysis might differ for CHIM studies conducted in LMICs. The main deliverable will be a resulting white paper. Project researchers are not expected to complete the white paper in the 8-week project period, and are welcome, but not expected to continue throughout the writing process alongside other co-authors. All work will be acknowledged through authorship. A list of experts to consult will be included in project briefing materials. Experts will be introduced and lined up in advance of the project start to minimise delay.

The main questions of the project: - Which scaffolding methods and affordances are effective for an AI model to acquire knowledge about biosecurity threats for a model with no such knowledge? - Whether such methods might be used to create a useful behavioral benchmark for AI models? As there are many tools that might improve performance of an AI model, there is a room for participants to choose which methods do they want to use in their research. The result of the research will be summarised into a report, which will be shared with experts and key stakeholders. Experts will be consulted before making any output public, and anything that might be an infohazard will be removed before publication.

Over the seven weeks, a researcher will contribute to answering some of these questions. Questions one, two, and three deal with specific countries, while questions four and five regionally. 1. What are the existing governance practices for managing the operation of high-containment level labs? Focusing on a specific country and understanding the context, such as which government institutions and their governance responsibility and mandates. If multiple institutions are involved, what governance responsibility exists within each? 2. What gaps exist in the current governance structure of high-containment levels labs? This will involve identifying lacking regulations, policies, or administrative measures; also, the question will seek to understand gaps within the regulatory institution. Are the institutions making the governance lacking? Are their issues with centralization or decentralization? Are the institution's capacity or resources constraints to make more effective governance? 3. What actional opportunities are available to strengthen and make governance more robust and effective? Researchers will justify with evidence what makes the recommendation actionable and explain how the recommendation connects to one or several gaps identified. 4. What regional organizations/institutions could shape and improve the governance, and in what ways? What's currently lacking? 5. What actional opportunities are available for the identified regional organization/institutions to strengthen and make governance more robust and effective? Researchers will justify with evidence what makes the recommendation actionable and explain how the recommendation connects to one or several gaps identified.