Governing emerging technologies poses a distinctive challenge for innovation policy: how to provide direction amid deep uncertainty, where technological trajectories, societal impacts, and evaluative criteria remain unsettled. Artificial intelligence (AI) exemplifies this condition, as governments seek to steer its development toward societal goals amid rapidly evolving capabilities and contested policy objectives. Existing policymaking frameworks, which emphasize the coherence, consistency, credibility, and comprehensiveness of the policy mix, implicitly assume that these attributes can be jointly maximized—an assumption that strains under conditions of emergence. Drawing on Paradox Theory and interviews with 28 senior AI policymakers from 22 countries, this article conceptualizes emerging technologies governance as a system of enduring and interdependent tensions. We identify eight structural tensions that systematically shape the elements and processes of the policy mix. Rather than resolving these tensions, effective policymakers adopt paradoxical strategies that sustain competing demands in productive balance over time. The article advances innovation policy in three ways. First, it introduces Paradox Theory as a foundational theoretical logic for analyzing policy mixes under deep uncertainty. Second, it reconceptualizes directionality as the capacity to navigate persistent contradictions rather than to achieve static alignment. Third, it reframes the role of governments in transformative innovation policy as active managers of paradoxes, offering a theoretically grounded alternative to coherence-centered models of policy design.