According to two sources familiar with classified intelligence estimates, the National Security Agency told lawmakers that it is spending billions of taxpayer dollars this year evaluating and testing advanced artificial intelligence models. The price tag, described as significantly greater than previously known, has led those sources to say that a more comprehensive AI regulatory system could cost the government tens of billions of dollars per year. The sources added that after a series of high‑profile hacking and security incidents, the NSA’s Artificial Intelligence Security Center began testing frontier models to identify potential national‑security vulnerabilities. The high cost of those efforts is said to be bringing new urgency to the debate about the government’s role in reviewing AI models designed by some of the most powerful and resourceful companies in the world, and may intensify calls for the frontier labs to shoulder the financial burden of reviewing the models as spending is expected to balloon with technological advances. Estimates of prior proposals for federal AI oversight are far lower: the Congressional Budget Office estimated the AI Security and Innovation Act, a bipartisan House proposal to establish a center on AI risks and facilitate mitigation, would cost roughly $20 million per year, while a separate House bill to create a reporting and tracking system for AI calls for $36 million in total over the next five years, according to the CBO. Some tech leaders, such as Elon Musk, have suggested that frontier labs review each other’s models before release without government oversight. The Information reported that Google, OpenAI and Anthropic are jointly working on a plan to create their own AI safety‑focused “standards body,” an idea that some AI safety experts have rejected as effectively allowing the firms to police themselves. An alternative discussed in the source is for the government to levy a tax on AI companies to fund safety operations, ensuring independence and preventing the financial burden from falling primarily on taxpayers. Anthropic and OpenAI have indicated they want greater federal oversight and may be open to paying for it. The NSA’s AI testing efforts are currently funded by classified portions of the federal national security budget, the two sources said, though the exact dollar amount spent so far is not clear. The Pentagon declined to comment, with a Defense Department spokesperson stating that for security reasons the department does not discuss the technical architecture or resource allocation for its AI tools. The Defense Department’s annual budget is close to $1 trillion, and some budgetary funds already allocated to the military could be shifted into AI‑related spending, potentially reducing the apparent price tag. One source said the biggest AI‑related cost to the NSA thus far has been the extra computing power needed to run and test the models on chips, noting that purchasing computing power has become extremely expensive worldwide amid a surge in demand and insufficient chip production. As an example, The Information reported that Anthropic has lined up computing deals that could cost as much as $517 billion. Another major expense is personnel: some top AI engineers have been offered salaries in the hundreds of millions of dollars, far outpacing what the government can match. Although the NSA is widely regarded as having the deepest roster of technical experts within the government, it may still be unable to compete with the giant pay packages offered by frontier labs. Nathan Calvin, general counsel at the AI advocacy organization Encode, was quoted as saying, “Having in‑house AI evaluation capability I think is extremely important and necessary. But it is genuinely expensive… You’re competing in bidding with some of the most price‑insensitive customers.” Experts such as Nat Purser, director of U.S. Policy at the AI Verification and Evaluation Research Institute, have urged lawmakers to strengthen third‑party audits of AI models, noting that people often掉估
