AgentsCoSAI MCP Security White Paper: 12 Threat CategoriesCoSAI·high signalXBlueskyLinkedInCopy linkFirst comprehensive MCP security taxonomy with 40 attack vectorsSourceSource pageCoSAI↳ Follow the threadStack layer / Threat patternAUTOSIGMA Turns Unstructured Threat Reports Into Sigma Detection Rules Without Trusting the LLM AlonearXiv 2608.19011Stack layer / Threat patternCatastrophic Learning: Poisoned Data Can Block a Continual-Learning Model From Acquiring Knowledge It Has Not Seen YetarXiv 2608.18976Stack layer / Threat patternSecrets Sitting in an Agent's Context Leak Through Ordinary Replies, and Stronger Models Leak MorearXiv 2608.19857Threat pattern / Update threadMicrosoft's Skala 1.1 DFT functional hits 2.8 kcal/mol on GMTKN55 and gold-medals 32 of 55 categoriesMicrosoft Research BlogStack layer / Threat patternSupernova Launched at Product Hunt #2 Selling the BI Stack as an MCP Endpoint Instead of a DashboardProduct HuntThreat pattern / ContrastOWASP Top 10 for LLM Applications barely agrees with 7,714 real incidentsarXivStack layer / Threat patternFortinet Buys Virtue AI to Add Runtime Protection for Agents and MCP ToolsFortinetPolicy dependency / Stack layerLet the model pick NoThink, Short, or Long at response start and cut mean tokens 41% for a 1.4 point accuracy lossarXiv 2608.20256