RAND Securing AI Model Weights and the SL3 framework
Preventing Theft and Misuse of Frontier Models (2024); Achieving AI Model Weight Security Level 3 (2026)
The reference framing for protecting model weights against theft, from opportunistic attackers at SL1 to top-priority nation-state operations at SL5, adopted by frontier safety frameworks and the EU code of practice.
Contribute
Spot an error, or know a record we should add? Members can propose additions and submit corrections.