What we do
01 / DETECT
02 / REPORT
03 / MITIGATE
It takes an intersectional community of dedicated developers and scientists to identify and confirm relevant jailbreaks and errors of commission. We seek to limit the spread of bioweapons by enabling scientifically helpful AI without risk. Our goal is to raise awareness and foster understanding of this issue.
Upon detection, we will endeavor to report the issues observed through community awareness, direct communications, and (where applicable) consulting services. We prioritize the security of the information we provide, ensuring that it is reliable, accurate, and protected from misuse. We also keep a log of jailbreaks and breaches.
Awareness is not enough. Through our community and more technologically sophisticated approaches, we equip AI-development companies with knowledge of what needs to be fixed, and federal agencies and sovereign actors with the information they need to make specific preparations.
Our Community
Scientists
Bench-experienced practitioners of molecular biology, geneticists, biochemistry, chemical engineering, synthetic biologics, translational medicine researchers, infectious disease specialists, and more.
Through hands-on experience and real world experiences, biological scientists provide the domain and subject matter expertise required to assess the validity and precision of commissions that may not be biosecure.
Software Engineers
Call the white hats, gray hats, red hats, or whatever else, the software engineering skillset equips an individual with the sensibilities and skillsets needed to provide the most cutting edge insights around interacting with publicly-accessible LLM's and Ai systems.​
​
Beyond clever elicitations and prompting, vulnerabilities such as code injection provide computation-native specialists uniquely potent capabilities for information extraction.
AI Researchers
Who better to find vulnerabilities than the mathematicians, computer scientists, statisticians and physicists responsible for the amazing AI models and LLM implementations that we have all come to know and love.
​
The intuitive understanding of the relationship between a model's architecture and the nuances of its responses equips these individuals to discover issues before they become apparent to general users.