AIToday
AI Safety & AlignmentAI Regulation & PolicyAI Business & IndustryTHE DECODERPublished: Sep 25, 2026, 22:00 JST

NSA tells Congress: billions this year testing AI models

NSA tells Congress: billions this year testing AI models

3 Key Points

  1. What happened

    The NSA told Congress it will spend billions of dollars this year testing advanced AI models, with computing power the biggest expense, The Washington Sun reports. Lawmakers now expect full-scale oversight to cost tens of billions of dollars a year, against a Congressional Budget Office estimate of about $20 million a year.

  2. Why it matters

    The gap between the roughly $20 million earlier estimate and the tens of billions now expected signals that the cost of watching over AI is landing on the government budget, not just on private labs.

  3. What to watch

    The bill hinges on who pays, with Nat Purser of the AI Verification and Evaluation Research Institute proposing a levy on developers for independent testing. The article reports developers have signaled they might be willing to pay for more government oversight.

WHO IT HITSUS government budget staff and lawmakers weighing AI oversight funding, plus hospital finance and insurer claims teams, both of which the article says use AI to fight over billing and have added nearly $1 billion in costs over two years.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Context & Analysis

The NSA's disclosure to Congress marks a shift in how the cost of AI oversight is being tallied. The agency says computing power is its biggest expense, and staffing adds to the bill because it has to compete with the pay packages AI labs offer. That has pushed lawmakers' expectations for full-scale oversight to tens of billions of dollars a year, far above the Congressional Budget Office's earlier estimate of about $20 million a year for a bipartisan bill to create a center for AI risks. Whether that funding materializes — and who covers it — is likely to shape how quickly oversight takes shape.

In health care, the cost driver is different: not the price of AI itself, but how hospitals and insurers use it for billing. The Blue Cross Blue Shield Association accuses hospitals of using AI-assisted coding to code cases as more severe and add secondary diagnoses, collecting nearly $12,000 more per case on average without any apparent change in treatment. For hospitals it pays off — McLaren Health Care CFO Dave Mazurkiewicz says the system brings in an extra $1 million a month using software from SmarterDx, and the vendor gets a cut of the added revenue. Insurers respond by using AI to comb through records for claims they can deny, which Caroline Pearson of the Peterson Health Technology Institute says leads both sides to go several rounds per dispute "because it's cheap."

Both cases point to the same tension: the money AI saves in some places shows up as new costs elsewhere. David Brailer, writing in Health Affairs, says he expects AI to raise overall health care spending, just as electronic health records did before it. For the NSA, the debate over who should foot the bill has already started, with Nat Purser of the AI Verification and Evaluation Research Institute proposing a levy that would have developers help pay for independent testing — and with Anthropic and OpenAI having signaled they might be willing to pay for more government oversight, according to The Washington Sun.

FAQ
Why is the NSA spending so much on AI testing?
Computing power is the biggest expense, according to a source cited by The Washington Sun. Staffing also adds to the bill because the NSA has to compete with the pay packages AI labs offer.
How is AI raising costs in US health care?
The Blue Cross Blue Shield Association says hospitals used AI-assisted coding to run up nearly $1 billion in extra costs over two years, collecting nearly $12,000 more per case on average. Insurers, meanwhile, use AI to comb through records for claims they can deny.
Are AI companies willing to pay for government oversight?
According to The Washington Sun, Anthropic and OpenAI have signaled they might be willing to pay for more government oversight.

Get the latest AI Safety & Alignment news every morning

For example, today's edition would include:

  • White House tells OpenAI, Anthropic: hold models from UK testersTHE DECODER · 3h ago
  • NYC's Julie Menin unveils AI kill-switch bills, $25,000 finesFortune AI · 3h ago
  • Pentagon seeks $30.3 million for AI lie detector "Polygraph+"MIT Technology Review AI · 3h ago

AI-summarized, only the topics you pick: one digest a day via Email, LINE, or Slack.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleMicrosoft's Copilot super app targets business users