The next frontier of AI-enabled influence is not content. It is cognition. As AI systems become primary mediators of individual judgement, institutional fact-finding, and collective decision-making, two converging threat vectors are emerging: the targeted manipulation of cognitive and emotional states at the individual level, and the systematic contamination of the data ecosystems through which societies construct authoritative knowledge. Drawing on findings from a dedicated threat casting workshop, this panel presents adversarial scenarios projected to 2036, maps the governance frameworks that do not yet exist, and advances a policy agenda for what states and institutions must prioritise before the window for effective action closes.