IARPA posts LLM RFI
The Intelligence Advanced Research Projects Activity (IARPA) has issued a request for information (RFI) on characterizing large language model (LLM) biases, threats, and vulnerabilities. They seek established characterizations of potential vulnerabilities and threats that could impact the secure use of LLMs by intelligence analysts.
While LLMs hold significant promise in transforming various sectors, their human-like interaction capabilities have raised concerns about erroneous and potentially harmful behavior, posing risks to end-users.
The RFI aims to gather frameworks to categorize and characterize these vulnerabilities and threats associated with LLM technologies, including both "white box" models (with privileged access to parameters or code) and "black box" models (without privileged access). The deadline for responses to the RFI is August 21, 2023.