AI Costs Surge for NSA, Hospitals and Insurers as Oversight Price Tag Grows
The NSA is spending billions to test advanced AI models, while lawmakers warn that full-scale oversight could cost tens of billions annually. In US healthcare, AI-assisted billing codes have already added nearly $1 billion in costs over two years.
The National Security Agency has already spent billions of dollars testing advanced artificial intelligence models, with most of the money going toward computing power, according to reporting by The Washington Sun. The figure underscores how the race to adopt cutting-edge AI is straining budgets far beyond initial projections.
Lawmakers now expect that full-scale oversight of AI across the federal government could cost tens of billions of dollars a year. That estimate stands in stark contrast to earlier projections from the Congressional Budget Office, which had put the figure at just $20 million. The gap between the two numbers highlights how quickly the financial reality of AI governance has outpaced official forecasts.
The NSA's spending is largely driven by the enormous computational resources required to train and run advanced AI models. These systems demand specialized hardware, vast data storage, and significant energy consumption, all of which carry steep price tags. For an agency whose core mission involves signals intelligence and cybersecurity, the ability to test and deploy AI is seen as increasingly essential — but also increasingly expensive.
The cost overruns are not limited to the intelligence community. In the US healthcare sector, AI-assisted billing codes have driven up costs by nearly $1 billion over two years. The use of algorithms to assign medical codes — which determine how much hospitals and insurers get paid — has introduced new layers of complexity and expense. While such tools are often marketed as efficiency drivers, the early evidence suggests they can also inflate administrative costs.
Hospitals and insurers are now grappling with the dual pressure of adopting AI to stay competitive while managing the unexpected financial fallout. The billing code issue is particularly sensitive because it touches on reimbursement rates, compliance risks, and the potential for coding errors that can lead to audits or denied claims. For patients, the downstream effect may be higher premiums or reduced services if providers pass on the costs.
The broader pattern is one of sticker shock. From national security to healthcare administration, organizations that rushed to integrate AI are discovering that the technology is not a one-time investment but an ongoing operational expense. Computing power, data management, and specialized talent all require sustained funding, and the returns are not always immediate or guaranteed.
Lawmakers are now faced with a difficult choice: authorize massive new spending to oversee AI effectively, or risk underregulation as the technology spreads. The Congressional Budget Office's earlier $20 million estimate now looks wildly optimistic, and the revised figure of tens of billions annually suggests that AI oversight could become one of the fastest-growing line items in the federal budget.
For the NSA, the challenge is compounded by the classified nature of its work. Much of the spending on AI testing is shielded from public view, making it hard to assess whether the investments are yielding actionable intelligence or simply consuming resources. The agency has not publicly detailed its AI budget, but the reported billions indicate that AI has become a central — and costly — component of its operations.
In healthcare, the nearly $1 billion increase in costs tied to AI-assisted billing codes over two years is a warning sign for an industry already struggling with rising expenses. Insurers may respond by tightening reimbursement rules, while hospitals could face pressure to justify the use of automated coding systems. The financial burden ultimately falls on the broader system, including taxpayers who fund public health programs.
The convergence of these cost pressures — in intelligence, healthcare, and beyond — points to a larger question: who pays for the AI revolution, and how will those costs be managed? As agencies and companies scale up their AI capabilities, the price of intelligence, in every sense, is becoming impossible to ignore.
6
