An essay by Jun-Yong Park argues that two concurrent developments create a strategic vulnerability: artificial intelligence systems now produce genuine research-level mathematics while the United States weakens the institutional pipeline that produces humans capable of verifying such work. The author contends that mathematical capacity — the trained ability to verify, interpret, and challenge reasoning — functions as infrastructure built over generations, not a byproduct of theorem production.

The essay cites the May 2026 AI disproof of a longstanding Erdős conjecture on the planar unit distance problem as evidence of AI's advancing mathematical capability. Simultaneously, recent disruptions to federal support for the mathematical sciences threaten the human expertise needed to audit these results. Park proposes treating mathematical capacity as a strategic asset comparable to semiconductor capability.

To address the opacity of AI reasoning, the essay recommends that systems performing consequential reasoning be required to expose their decision-critical claims in formal, machine-checkable form. This would convert part of AI reasoning from opaque persuasion into auditable structure, enabling human verification without requiring every validator to replicate the entire derivation.

The proposal reframes AI safety as a verification infrastructure problem rather than a model alignment problem alone. If adopted, it would shift regulatory focus toward formal proof standards and tooling for machine-checkable reasoning, creating demand for proof assistants and formalization frameworks that bridge AI output and human auditability.