NIST CAISI signs pre-deployment evaluation agreements with major AI labs
NIST's Center for AI Standards and Innovation announced formal agreements with Google DeepMind, Microsoft, and xAI to evaluate frontier models before public release. The shift represents government oversight of model launches and has already completed 40+ evaluations including unreleased state-of-the-art systems.
The Center for AI Standards and Innovation (CAISI) at NIST announced new agreements enabling government evaluation of AI models before public availability, alongside post-deployment assessment and targeted research. The agreements with Google DeepMind, Microsoft, and xAI formalize a process CAISI has already completed more than 40 times, including evaluations of unreleased frontier models.
Pre-deployment assessment framework
The agreements represent a formal shift toward government oversight of model releases before public availability. Pre-deployment evaluations and targeted research to assess frontier AI capabilities and advance AI security establish a structured process where national security testing occurs before deployment.
Scope and precedent
The evaluations cover frontier AI capabilities and security assessment. This arrangement allows NIST researchers direct access to models still in development, enabling assessment of risks before they reach the public. The participation of three major labs signals industry acceptance of pre-release evaluation as standard practice.