Governing frontier general-purpose AI in the public sector: adaptive risk management and policy capacity under uncertainty through 2030

Governments face an "evidence dilemma" where AI capabilities are advancing rapidly but knowledge about potential harms and effective safeguards remains incomplete, requiring policy decisions under significant uncertainty rather than waiting for perfect information.

Static compliance models are insufficient for governing frontier AI in the public sector; instead, adaptive risk management and scenario-aware regulation that can adjust to multiple plausible futures through 2030 are needed to keep pace with technological change.

AI adoption outcomes in government depend heavily on organizational and institutional factors—including data arrangements, accountability structures, and organizational routines—not just technical model performance, meaning governance must address sociotechnical transformation alongside technical safeguards.

Effective public-sector AI governance requires capability monitoring, risk tiering, and conditional controls that allow governments to learn and adjust policies as evidence about AI impacts accumulates, rather than implementing fixed rules upfront.

Public values, institutional dynamics, and data collaboration capacity are central to responsible AI deployment in government, suggesting that governance frameworks must integrate considerations of democratic accountability and organizational readiness alongside technical risk assessment.