Definition
The set of technical and procedural controls used to protect an AI model's trained parameters (its 'weights') from theft, tampering, or leakage — much like how a company protects its most valuable source code. A new framework called SL3 lays out 262 specific controls organizations can adopt within six to twelve months.
Why it matters
A company's model weights represent years of investment and competitive advantage; if stolen, a rival or adversary can replicate or misuse the model without any of the original training cost, so boards should know whether this asset class is protected like other crown-jewel IP.