Abstract
The Responsible Research and Innovation (RRI) framework is essential for navigating the Collingridge dilemma, yet it struggles with a core tension: deliberative depth versus operational agility. This Perspective argues for reframing AI from a governance object to a governance instrument to bridge this gap. We illustrate how agentic AI can augment the four RRI pillars—anticipation, reflexivity, inclusion, responsiveness—by automating burdens, thereby creating space for human deliberation. Crucially, we advance ‘using AI to govern AI’ as a meta-governance framework to mitigate risks like accountability obfuscation. We conclude by calling for Governance Engineering: designing scalable human-AI ecosystems to transform this tension into adaptive solutions.