Sources: the NSA told lawmakers it is spending billions this year to test AI models; proposals for a US AI regulatory body estimated costs of $20M-$40M per year

The NSA's billions spent on AI testing means the government will likely demand AI developers fund future regulatory evaluations.

The National Security Agency (NSA) has disclosed to lawmakers that it is currently spending billions of dollars on testing advanced artificial intelligence models this year. These evaluations are being conducted by the NSA’s Artificial Intelligence Security Center to identify potential national security vulnerabilities. The significant cost of these AI testing efforts, which far exceeds previous estimates for government AI oversight, is reportedly driving discussions about the government's role in regulating AI. Lawmakers are now considering that a comprehensive AI regulatory system could cost tens of billions of dollars annually. The NSA's current AI testing is funded through classified national security budgets, with computing power and specialized personnel being the primary expenses. This revelation comes amid broader debates on AI regulation, with some proposals for federal oversight costing significantly less, around $20 million to $40 million per year.

The sheer scale of the NSA's AI testing expenditure, running into billions, underscores the immense computational and human resources required for robust AI evaluation. This dwarfs prior estimates for federal AI regulatory bodies, which were in the tens of millions annually, suggesting that any government-led oversight on a similar scale would be prohibitively expensive for taxpayers. Consequently, there is growing momentum for AI frontier labs to shoulder the financial burden for these critical safety and security assessments, potentially through taxes or levies.

This development signals a significant shift in the AI regulatory landscape, moving from theoretical discussions to substantial government investment in AI security. The high cost implies that purely government-funded oversight may be unsustainable, pushing the responsibility and cost onto the very companies developing these advanced models. It also raises questions about the feasibility of industry-led self-regulation versus independent government or third-party audits, particularly as AI capabilities continue to advance rapidly and the demand for computing power soars.