A formal model of AI integration into public administration highlights how compliance mechanisms intended to ensure reviewability can create stable approval boundaries that political successors learn to navigate while maintaining the appearance of lawful administration.

The arXiv paper by Peterson (arXiv:2604.21103) develops a formal model examining institutions' choices over scale of automation, degree of codification, and safeguards on iterative use, demonstrating that systems improving oversight initially can later heighten vulnerability to strategic internal exploitation and that AI expansions become difficult to unwind once embedded. Original coverage of this work overlooks its direct implications for real-world regulatory persistence, such as how the EU AI Act's risk-based framework (artificialintelligenceact.eu) aims to impose durable rules transcending national elections yet may still permit interpretive navigation of its compliance surface as described in the model.

Synthesis with the Biden 2023 Executive Order on Safe, Secure, and Trustworthy AI (whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence), which drove codification in federal agencies, reveals patterns where algorithmic tools in benefits and criminal justice have shown similar lock-in across U.S. administrations. What prior analyses missed is the model's prediction that making AI usable for cheaper, scalable decisions teaches future governments to exploit the compliance layer without overt violation.

The key gap in designing durable compliance systems is the lack of mechanisms like mandatory recalibration triggers tied to electoral cycles or adaptive thresholds responsive to both technological advance and global regulatory shifts, leaving governance frameworks vulnerable to surviving political turnover in ways that entrench rather than prevent administrative capture.