grantmaking.ai Launch Round
$40,000
$100,000
Collaborations with Biosecurity insitutions for expert domain knowledge
GPU and LLM costs
Funding research students wokring on the project
BMG is a drop in open ai compatible gateway that screens LLM and Agent calls for dual use biological risk.
BMG is a drop in open ai compatible gateway that screens LLM and Agent calls for dual use biological risk.
Project Details
Updated 07/12/26 · Provided via application · VerifiedCurrently all dual use risk for LLMs and Agents is delegated at the provider level . What we see is frequent jailbreaks with Fable 5 being jailbroken in less than a day. We believe that LLMs and Agents should have a tailored moderation gateway specificallly designed for the biological domain . This gateway will block dual use requests in research environments , bioinfromatics agents such as Biomni and KDenseAnalyst will still enabling legitimate research. We will implement a scalable architecture that combines keyword search , traditional ML and policy based safeguards so that no significant delay is observed when using LLMs or Agents through the gateway when no significant risk is observed. Our proposition in one line is that Biosafety / Biosecurity is incredibly important based on findings from our
paper BioVeil Matrix: https://arxiv.org/html/2605.00927v1 .
and the BioVeil Taxonomy Website: https://bioveilmatrix.com/
It needs a dedicated security layer.
Theory of Impact
Updated 07/12/26 · By grantmaking.aiIt reduces biological risk for all models / LLM providers horizontally and provides a single point of safeguard for biological risk. It could cover safety on flagship to smaller models seamlessly
People
Updated 07/12/26 · By grantmaking.aiTeam Member
Funding Asks
Applied
Discussion
Sign in to comment
No comments yet. Be the first to share your thoughts.