Published research from our lab
Peer-style academic outputs from our researchers, surfacing new attack classes and defensive primitives across modern AI systems.
Safety, Security, and Cognitive Risks in State-Space Models: A Systematic Threat Analysis with Spectral, Stateful, and Capacity Attacks
A systematic threat analysis of state-space models introducing spectral, stateful, and capacity attack classes that expose new safety, security, and cognitive risks unique to SSM-based architectures.
Safety, Security, and Cognitive Risks in World Models
A structured examination of how world models introduce a distinct class of safety, security, and cognitive risks — and what defensive primitives are needed to govern them in real-world deployments.
More research from our team is in active preparation. Check back for new publications.
Patents from our lab
Filed patents covering novel approaches to AI security, threat detection, and defensive primitives.
Patent Application IN 2*2****5*6*6
A patent application from SovereignAI Security Labs filed in the India Patent Office covering proprietary innovations in AI security and threat intelligence.
Additional patent filings are in progress as our research matures into protected IP.