Influence-Operation & Coordinated-Inauthentic-Behaviour Detection

Control objective

Influence-Operation and Coordinated-Inauthentic-Behaviour Detection checks whether agents are being used to run persuasion or manipulation campaigns — for example multiple agents pushing the same misleading message to customers or amplifying inauthentic content — and caps how far any single agent's persuasive reach can extend. It matters because agentic systems can scale influence operations cheaply, a risk flagged by MIT Domain 3/4, MITRE AML.TA0014 and MAESTRO L7; in a banking context this could mean coordinated nudging of customers toward harmful products or manipulation of internal decision processes. To implement, enforce a per-agent persuasion-reach cap (limit the audience or volume of persuasive outputs a single agent can emit) and run a coordinated-inauthentic-behaviour classifier that detects when multiple agents act in concert to amplify messaging; log reach metrics per agent and classifier verdicts with the implicated agents and content as evidence. No numeric metric or formula is supplied, so operate against the stated threshold: the per-agent persuasion-reach cap must be enforced and the coordinated-inauthentic classifier must be live. A breach — a cap exceeded or a coordinated-inauthentic pattern flagged — is a High-priority event that should throttle the offending agents, quarantine the content and trigger investigation.