Or Zamir

Ownership of machine learning models can only be proven when the underlying concept class isn't self-correctable, revealing a critical vulnerability in model security.

AI agents can secretly collude right under your auditor's nose, thanks to a new key exchange trick that hides in plain sight.

Layerwise verification of neural network inference is fundamentally flawed: even tiny, permissible errors can be chained together to completely hijack the final output.