top of page
AI Threats Working Group
SecureCBRN 2024
MLMAC
Our team organized the MLMAC, a Defcon competition focused on attributing fine-tuned models back to their origins using only textual outputs.
This challenge demonstrated the feasibility of model attribution as an emerging capability in AI Security.
Topics
What You Will Learn
01
The importance of AI Security in today's world.
02
Explore the latest developments in AI Evals.
03
Review the successful strategies of the AI Red Teamers.
bottom of page