A Research Agenda for AI Training Interconnect Security
Threat modeling the high-bandwidth communication layer connecting accelerators, hosts, and storage in distributed AI training.
Read research agendaWriting
Writing and research notes
Research notes and essays on AI security, AI systems, and the security implications of emerging capabilities.
Misalignment Contagion in Multi-Agent Games
A small pilot separating context-bound behavioral contagion from persistent trait contagion in repeated agent interactions.
Read findingsPersistent by Design, Vulnerable by Default
What query-only memory injection attacks reveal about the security boundary between an aligned model and its persistent memory.
Read research noteWhat China’s 3.15 GEO Exposé Reveals About AI Answer-System Security
A security analysis of fabricated online consensus and the risks it creates for AI systems that synthesize answers from public information.
Read essay