A critical finding from this year’s analysis is the vulnerability of the “connective tissue” of the contemporary AI economy. The rapid adoption of the Model Context Protocol (MCP) and agentic and agent-to-agent (A2A) protocols has created a vast and often unmonitored attack surface. As detailed in this year’s report, while threat actors continue to bypass generative AI model guardrails to elicit unsafe or compromising outputs, they are also compromising agents that have the authority to execute processes, access databases, and push code on behalf of humans.