Mitigating Risks from Dual-Use Research

A response to OSTP's request for information on developing a policy for research with certain pathogens
October 17th 2023

On October 11th, 2023, the Institute for Progress filed the following comment in response to the Office of Science and Technology Policy’s Request for Information on developing a Revised Policy for research with certain pathogens and toxins. OSTP requested comments on how updating research oversight policies could mitigate risks from dual-use research and enhanced potential pandemic pathogens (ePPP) while minimizing undue burden on institutions.


Recent advances in life science research, biotechnology, and artificial intelligence tools could have enormous medical and public health benefits, but these advancements also increase risks of misuse. We appreciate the opportunity to comment on the potential changes to the current Dual Use Research of Concern (DURC) and the Potential Pandemic Pathogen Care and Oversight (P3CO) frameworks for oversight and review of federally funded life science research on pathogens and toxins.

Overall, we strongly support the National Science Advisory Board for Biosecurity’s (NSABB) recommendations, and look forward to working with you to strengthen federal biorisk management. We are especially supportive of having one harmonized federal policy that applies to all federally funded research, and that takes into account current and future life science research on pathogens and toxins. Our research informs the following recommendations for a Revised Policy. 

Centralizing biosafety reporting

We strongly support a single Revised Policy. Current oversight of federally funded DURC/ePPP research is fractured, with biosafety and biosecurity left to the discretion of academic institutions. The patchwork of current policies is confusing and leaves large holes in implementation. One Revised Policy would ensure consistent compliance across institutions, and clarify institutional responsibility for identifying DURC/ePPP research. 

To foster consistent, comprehensive institutional reporting, NIH might consider implementing a grading or certification system to evaluate institutional adherence to reporting requirements. This certification could then serve as a merit indicator during grant applications and could be disclosed in public reports. Streamlining reporting through a user-friendly NIH portal can further promote timely, accurate submissions. 

For nuanced situations where guidance may be needed, the NIH could set up a dedicated helpdesk or hotline for institutions, allowing it to confer with domain experts. Regularly scheduled reviews of Institutional Biosafety Committees, paired with constructive feedback, will provide a framework for assessing adherence to the Revised Policy. Additionally, transparent spot audits — followed by a debriefing session and publishing de-identified findings in a public repository — can offer clarity on regulations and foster a culture of shared learning and best practices across institutions.

We believe institutions, rather than Principal Investigators, should be responsible for identifying DURC and ePPP research. This responsibility could also fall under the purview of Institutional Biosafety Committees. Often, the methodology a lab group proposes for identifying DURC and ePPP research is not ultimately followed. For institutions to take on the role of identifying DURC and ePPP research, they need a mechanism to support follow up and decision-making either informally, through department/lab meetings, or more formally. For institutions that require extra resources and assistance, the NIH could have them indicate the need for a line item in the grant-making process for DURC and ePPP oversight. 

A Revised Policy should also include a method for verification of consistency in identifying DURC and ePPP research across institutions, as definitions differ. To verify, the NIH could check a subset of reviews for completeness and publicly compile examples of strong institutional review processes.

We recommend the development of a digital risk assessment platform to streamline proposal evaluations and ensure consistent decision-making across Institutional Biosafety Committees. This platform should feature a questionnaire designed to assess dual-use risks. Given the success and functionality of the Dual-Use Quickscan tool by the Netherlands Biosecurity Office (NBO), we propose that NIH collaborate with NBO for the implementation of a similar tool. This approach will ensure that the platform aligns with global best practices and contributes to comprehensive adherence to the Revised Policy.

Defining clear risk thresholds

We believe that while the provided definition of “reasonably anticipated” does improve clarity, a Revised Policy or guidance document should provide:

  1. tools for researchers to assess a “reasonably anticipated” outcome, and
  2. examples or relevant metrics of “reasonably anticipated” minimum (R0s) thresholds for commonly researched respiratory pathogens.

A Revised Policy or guidance document should help researchers assess if a proposed protocol may have a “reasonably anticipated” outcome. A Revised Policy should develop a questionnaire framework to guide thinking about this likelihood. This would ensure consistency at the institutional level. At the institutional level, a Revised Policy could require these questionnaires for Principal Investigators (PIs) to think through why their research might result in a pathogen that is more pathogenic or transmissible. These explanations would then be shared with the Institutional Biosafety Committee (IBC) (or other institutional process) as part of the review process  and subsequent affirmation as to whether they believe the work should proceed or not based on the reasonably anticipated outcomes.

Complementing a Revised Policy, tools that can help clarify federal intent with the “reasonably anticipated” threshold include NIH webinars and trainings, institutional level workshops, and an online forum in which researchers can access NIH domain experts. To gather data on researcher/institutional compliance, a Revised Policy should consider a no-fault anonymous reporting system.

A Revised Policy could provide risk thresholds for commonly researched respiratory pathogens, such as for pathogen X “reasonably anticipated” is defined as “above Y risk threshold, as determined by Z.” Here, “reasonably anticipated” would apply at the institutional level, and threshold definitions could be used to verify compliance after experiments were conducted. 

For example, NIH could provide a threshold for receptor-binding affinities for different influenza strains. These definitions could be similar to the Federal Select Agent Program’s definition of virulent Newcastle disease virus, to clarify the intent of the “reasonably anticipated” threshold: 

“A virulent Newcastle disease virus (avian paramyxovirus serotype 1) has an intracerebral pathogenicity index in day-old chicks (Gallus gallus) of 0.7 or greater or has an amino acid sequence at the fusion (F) protein cleavage site that is consistent with virulent strains of Newcastle disease virus. A failure to detect a cleavage site that is consistent with virulent strains does not confirm the absence of a virulent virus.”

Where possible, we believe that the determination as to what is “reasonably anticipated” should be quantitative. One approach for making these determinations would be for NIH to build a dedicated in-house forecasting team that could assign probabilities to certain risk thresholds of pathogenicity or transmissibility being reached. This team would make these probabilistic assessments on the basis of relevant reference classes of past research and its consequences, as well as its own subjective probabilistic judgment, in line with general best practices from the field of human judgmental forecasting.

Allowing countermeasure development during emergencies

We support the NSABB’s recommendation to remove blanket research exclusions. With the growing number of BSL-3 and BSL-3+ facilities and the proliferation of biotechnology and AI tools, removing blanket research exclusions addresses increasing risks associated with laboratory accidents. 

However, a Revised Policy should maintain an exemption for countermeasure development targeted at an actively circulating pathogen during a U.S.-declared public health emergency. The exemption would only apply to research on the actively circulating pathogen or similar variants. This would support the rapid development of medical countermeasures during an ongoing crisis, such as supporting the 100-day goal outlined in the 2022 National Biodefense Strategy. Once a declared public health emergency expires the exclusion should be reimposed with public notice. This exclusion should not apply to research increasing transmissibility or virulence with no direct medical countermeasure purpose. 

Addressing in silico experiments

While current DURC and ePPP oversight policies do not currently encompass in silico experiments, the inclusion of these experiments is becoming increasingly necessary. Rapid advancements in both AI and biotechnology pose new biosecurity risks. For example, it was recently demonstrated that a single neural network can be used to model a wide array of biomolecular assemblies. If made widely accessible, such developments could lead to high-risk applications like pathogen engineering or toxin production. Additionally, large language models are making these biological design tools accessible to a wider range of actors. Given these developments, we believe a Revised Policy should specify which computational tools — such as algorithms that can be reasonably anticipated to be used for predicting pathogen virulence or transmissibility — are subject to public accessibility restrictions in line with their security risks. In particular, reconsideration of the open-sourcing of state-of-the-art biological design tools is warranted.

Monitoring computational models

We agree with the proposed categorization of in silico research — specifically, the mention of computational models that could result in the design of enhanced pathogens or toxins. It will be crucial to manage the public accessibility of these rapidly advancing computational tools based on their potential security risks. To address this, we recommend commissioning a NSABB study or convening a focused workshop that includes relevant experts in industry and academia. This workshop could be reconvened regularly, perhaps annually or biannually, given the rapidly evolving nature of the field. This process would aim to identify safeguards that balance biosecurity risks with the need for continued innovation.