Open access
Sep 2026
Trust inversion and its exploitation: A threat model of the Model Context Protocol for agentic AI
The analysis groups risks into descriptor and capability drift, cross-tool trust propagation, and sampling-mediated server influence, while separating protocol, ecosystem, implementation, and deployment weaknesses.
M. Q. Sambo, Augustine Yusuf, I. Nwokoro et al.
· Journal of the Nigerian Soci... · 0 citations