The National Security Agency is allocating billions of dollars this year to evaluate and test cutting‑edge artificial intelligence models, according to two sources with access to classified intelligence estimates. The NSA’s newly emphasized AI Security Center runs frontier models on government hardware to uncover national‑security vulnerabilities that emerged after a series of high‑profile hacking incidents. Agency officials say the effort’s cost far exceeds earlier expectations, prompting lawmakers to warn that a full‑scale federal AI oversight regime could run into the tens of billions annually. The biggest expense cited is the acquisition of specialized computing power needed to run the models; global chip shortages have driven prices up, with one source noting that Anthropic alone has secured compute contracts that could reach $517 billion. Personnel costs also loom large, as top AI engineers command salaries in the hundreds of millions, far beyond what the government can match despite the NSA’s deep technical talent pool. Funding comes from classified portions of the national‑security budget, though the exact figure remains undisclosed. Experts argue that the government must invest in both hardware and expertise, and suggest mechanisms such as an AI‑company tax or required third‑party audits to shift the financial burden away from taxpayers while ensuring independent safety assessments.