The Centre for Long-Term Resilience recently contributed to the National Institute of Standards and Technology’s Safety Considerations for Chemical and/or Biological AI Models. Our response draws on our previous work and recent insights to address the National Institute of Standards and Technology’s queries regarding existing and emerging evaluation methods and their application to chem-bio AI models, as well as current and potential future mitigations for the misuse of these models.
In our response, we:
-
- Highlight gaps in existing and emerging evaluation methods of chem-bio AI models
- Argue that non-interactive risk assessments could be used much more widely, with minimal costs, and are likely the best first-line evaluation method for novel chem-bio AI models
- Detail potentially useful evaluation strategies that warrant further investigation
- Recommend that government assistance and guidance play a role in the development of these evaluations alongside technical and biosecurity experts
To read the full response, click on the ‘Download’ button below. If you would like to discuss our response in more detail, please get in touch with biosecurity@longtermresilience.org
