TLDR Tech / InfoSec
What We Should Do Instead of Trusted Cyber Access Programs and Open Weight Restrictions
It is pivotal to provide defenders with access to the most powerful, cyber-capable models, whether open-weight or frontier models. Instead of restricting access to these models, we should shift the paradigm to focus more on restricting attackers' access. Frontier labs should implement low-friction, cross-lab Know Your Customer (KYC) programs, do more to detect and block attackers, and publish time-to-detection-and-response metrics. Victims of AI cyberattacks should also be encouraged to share more detailed information about the attack and the attackers.
Read the full story