Apparently NIST needs all the help it can get.

U.S. AI Industry capability investment ($120 billion p.a.) is outpacing U.S. Government regulatory investment ($10 million) by +12,000x


THE WASHINGTON POST. This agency is tasked with keeping AI safe. Its offices are crumbling.

Funding challenges at the National Institute of Standards and Technology could jeopardize the Biden administration AI work.

NIST is one of the federal government’s oldest science agencies — with one of the smallest budgets. Initially called the National Bureau of Standards, it began at the dawn of the 20th century, as Congress realized the need to develop more standardized measurements amid the expansion of electricity, the steam engine and railways.

During his State of the Union address on Thursday, Biden called on Congress to “harness the promise of AI and protect us from its peril.”

“It is a hard problem,” said Tabassi, who was recently named the chief technology officer of the AI Safety Institute. “We don’t know quite how to evaluate AI.”

By Cat Zakrzewski

March 6, 2024 at 7:00 a.m. EST

At the National Institute of Standards and Technology — the government lab overseeing the most anticipated technology on the planet — black mold has forced some workers out of their offices. Researchers sleep in their labs to protect their work during frequent blackouts. Some employees have to carry hard drives to other buildings; flaky internet won’t allow for the sending of large files.

And a leaky roof forces others to break out plastic sheeting.

“If we knew rain was coming, we’d tarp up the microscope,” said James Fekete, who served as chief of NIST’s applied chemicals and materials division until 2018. “It leaked enough that we were prepared.”

NIST is at the heart of President Biden’s ambitious plans to oversee a new generation of artificial intelligence models; through an executive order, the agency is tasked with developing tests for security flaws and other harms. But budget constraints have left the 123-year-old lab with a skeletal staff on key tech teams and most facilities on its main Gaithersburg, Md., and Boulder, Colo., campuses below acceptable building standards.

Interviews with more than a dozen current and former NIST employees, Biden administration officials, congressional aides and tech company executives, along with reports commissioned by the government, detail a massive resources gap between NIST and the tech firms it is tasked with evaluating — a discrepancy some say risks undermining the White House’s ambitious plans to set guardrails for the burgeoning technology. Many of the people spoke to The Washington Post on the condition of anonymity because they were not authorized to speak to the media.

Even as NIST races to set up the new U.S. AI Safety Institute, the crisis at the degrading lab is becoming more acute. On Sunday, lawmakers released a new spending plan that would cut NIST’s overall budget by more than 10 percent, to $1.46 billion. While lawmakers propose to invest $10 million in the new AI institute, that’s a fraction of the tens of billions of dollars tech giants like Google and Microsoft are pouring into the race to develop artificial intelligence. It pales in comparison to Britain, which has invested more than $125 million into its AI safety efforts.

The cuts to the agency “are a self-inflicted wound in the global tech race,” said Divyansh Kaushik, the associate director for emerging technologies and national security at the Federation of American Scientists.

Some in the AI community worry that underfunding NIST makes it vulnerable to industry influence. Tech companies are chipping in for the expensive computing infrastructure that will allow the institute to examine AI models. Amazon announced that it would donate $5 million in computing credits. Microsoft, a key investor in OpenAI, will provide engineering teams along with computing resources. (Amazon founder Jeff Bezos owns The Post.)

Tech executives, including OpenAI CEO Sam Altman, are regularly in communication with officials at the Commerce Department about the agency’s AI work. OpenAI has lobbied NIST on artificial intelligence issues, according to federal disclosures. NIST asked TechNet — an industry trade group whose members include OpenAI, Google and other major tech companies — if its member companies can advise the AI Safety Institute.

NIST is also seeking feedback from academics and civil society groups on its AI work. The agency has a long history of working with a variety of stakeholders to gather input on technologies, Commerce Department spokesman Charlie Andrews said.

AI staff, unlike their more ergonomically challenged colleagues, will be working in well-equipped offices in the Gaithersburg campus, the Commerce Department’s D.C. office and the NIST National Cybersecurity Center of Excellence in Rockville, Md., Andrews said.

White House spokeswoman Robyn Patterson said the appointment of Elizabeth Kelly to the helm of the new AI Safety Institute underscores the White House’s “commitment to getting this work done right and on time.” Kelly previously served as special assistant to the president for economic policy.

“The Biden-Harris administration has so far met every single milestone outlined by the president’s landmark executive order,” Patterson said. “We are confident in our ability to continue to effectively and expeditiously meet the milestones and directives set forth by President Biden to protect Americans from the potential risks of AI systems while catalyzing innovation in AI and beyond.”

NIST’s financial struggles highlight the limitations of the administration’s plan to regulate AI exclusively through the executive branch. Without an act of Congress, there is no new funding for initiatives like the AI Safety Institute and the programs could be easily overturned by the next president. And as the presidential elections approach, the prospects of Congress moving on AI in 2024 are growing dim.

During his State of the Union address on Thursday, Biden called on Congress to “harness the promise of AI and protect us from its peril.”

Congressional aides and former NIST employees say the agency has not been able to break through as a funding priority — even as lawmakers increasingly tout its role in addressing technological developments, including AI, chips and quantum computing.

After this article published, Senate Majority Leader Charles E. Schumer (D-N.Y.) on Thursday touted the $10 million investment in the institute in the proposed budget, saying he “fought for this funding to make sure that the development of AI prioritizes both innovation and safety.”

A review of NIST’s safety practices in August found that the budgetary issues endanger employees, alleging that the agency has an “incomplete and superficial approach” to safety.

“Chronic underfunding of the NIST facilities and maintenance budget has created unsafe work conditions and further fueled the impression among researchers that safety is not a priority,” said the NIST safety commission report, which was commissioned following the 2022 death of an engineering technician at the agency’s fire research lab.

Leaking ceilings

NIST is one of the federal government’s oldest science agencies — with one of the smallest budgets. Initially called the National Bureau of Standards, it began at the dawn of the 20th century, as Congress realized the need to develop more standardized measurements amid the expansion of electricity, the steam engine and railways.

The need for such an agency was underscored three years after its founding, when fires ravaged through Baltimore. Firefighters from Washington, Philadelphia and even New York rushed to help put out the flames, but without standard couplings, their hoses couldn’t connect to the Baltimore hydrants. The firefighters watched as the flames overtook more than 70 city blocks in 30 hours.

NIST developed a standard fitting, unifying more than 600 different types of hose couplings deployed across the country at the time.

Ever since, the agency has played a critical role in using research and science to help the country learn from catastrophes and prevent new ones. Its work expanded after World War II: It developed an early version of the digital computer, crucial Space Race instruments and atomic clocks, which underpin GPS. In the 1950s and 1960s, the agency moved to new campuses in Boulder and Gaithersburg after its early headquarters in Washington fell into disrepair.

Now, scientists at NIST joke that they work at the most advanced labs in the world — in the 1960s. Former employees describe cutting-edge scientific equipment surrounded by decades-old buildings that make it impossible to control the temperature or humidity to conduct critical experiments.

“You see dust everywhere because the windows don’t seal,” former acting NIST director Kent Rochford said. “You see a bucket catching drips from a leak in the roof. You see Home Depot dehumidifiers or portable AC units all over the place.”

The flooding was so bad that Rochford said he once requested money for scuba gear. That request was denied, but he did receive funding for an emergency kit that included squeegees to clean up water.

Pests and wildlife have at times infiltrated its campuses, including an incident where a garter snake entered a Boulder building.

More than 60 percent of NIST facilities do not meet federal standards for acceptable building conditions, according to a February 2023 report commissioned by Congress from the National Academies of Sciences, Engineering and Medicine. The poor conditions impact employee output. Workarounds and do-it-yourself repairs reduce the productivity of research staff by up to 40 percent, according to the committee’s interviews with employees during a laboratory visit.

Years after Rochford’s 2018 departure, NIST employees are still deploying similar MacGyver-style workarounds. Each year between October and March, low humidity in one lab creates a static charge making it impossible to operate an instrument ensuring companies meet environmental standards for greenhouse gases.

Problems with the HVAC and specialized lights have made the agency unable to meet demand for reference materials, which manufacturers use to check whether their measurements are accurate in products like baby formula.

Facility problems have also delayed critical work on biometrics, including evaluations of facial recognition systems used by the FBI and other law enforcement agencies. The data center in the 1966 building that houses that work receives inadequate cooling, and employees there spend about 30 percent of their time trying to mitigate problems with the lab, according to the academies’ reports. Scheduled outages are required to maintain the data centers that hold technology work, knocking all biometric evaluations offline for a month each year.

Fekete, the scientist who recalled covering the microscope, said his team’s device never completely stopped working due to rain water.

But other NIST employees haven’t been so lucky. Leaks and floods destroyed an electron microscope worth $2.5 million used for semiconductor research, and permanently damaged an advanced scale called a Kibble balance. The tool was out of commission for nearly five years.

An AI revolution

Despite these constraints, NIST has built a reputation as a natural interrogator of swiftly advancing AI systems.

In 2019, the agency released a landmark study confirming facial recognition systems misidentify people of color more often than White people, casting scrutiny on the technology’s popularity among law enforcement. Due to personnel constraints, only a handful of people worked on that project.

Four years later, NIST released early guidelines around AI, cementing its reputation as a government leader on the technology. To develop the framework, the agency connected with leaders in industry, civil society and other groups, earning a strong reputation among numerous partiesas lawmakers began to grapple with the swiftly evolving technology.

The work made NIST a natural home for the Biden administration’s AI red-teaming efforts and the AI Safety Institute, which were formalized in the November executive order. Vice President Harris touted the institute at the U.K. AI Safety Summit in November. More than 200 civil society organizations, academics and companies — including OpenAI and Google — have signed on to participate in a consortium within the institute.

OpenAI spokeswoman Kayla Wood said in a statement that the company supports NIST’s work, and that the company plans to continue to work with the lab to “support the development of effective AI oversight measures.”

Under the executive order, NIST has a laundry list of initiatives that it needs to complete by this summer, including publishing guidelines for how to red-team AI models and launching an initiative to guide evaluating AI capabilities. In a December speech at the machine learning conference NeurIPS, the agency’s chief AI adviser, Elham Tabassi, said this would be an “almost impossible deadline.”

“It is a hard problem,” said Tabassi, who was recently named the chief technology officer of the AI Safety Institute. “We don’t know quite how to evaluate AI.”

The NIST staff has worked “tirelessly” to complete the work it is assigned by the AI executive order, said Andrews, the Commerce spokesperson.

“While the administration has been clear that additional resources will be required to fully address all of the issues posed by AI in the long term, NIST has been effectively carrying out its responsibilities under the [executive order] and is prepared to continue to lead on AI-related research and other work,” he said.

Commerce Secretary Gina Raimondo asked Congress to allocate $10 million for the AI Safety Institute during an event at the Atlantic Council in January. The Biden administration also requested more funding for NIST facilities, including $262 million for safety, maintenance and repairs. Congressional appropriators responded by cutting NIST’s facilities budget.

The administration’s ask falls far below the recommendations of the national academies’ study, which urged Congress to provide $300 to $400 million in additional annual funding over 12 years to overcome a backlog of facilities damage. The report also calls for $120 million to $150 million per year for the same period to “stabilize the effects of further deterioration and obsolescence.”

Ross B. Corotis, who chaired the academies committee that produced the facilities report, said Congress needs to ensure that NIST is funded because it is the “go-to lab” when any new technology emerges, whether that’s chips or AI.

“Unless you’re going to build a whole new laboratory for some particular issue, you’re going to turn first to NIST,” Corotis said. “And NIST needs to be ready for that.”

Learn more:

Silicon Valley is pricing academics out of AI research.

With eye-popping salaries and access to costly computing power, AI companies are draining academia of talent.

As companies like Meta, Google and Microsoft funnel billions of dollars into AI, a massive resources gap is building with even the country’s richest universities. Meta aims to procure 350,000 of the specialized computer chips — called GPUs — that are essential to run the gargantuan calculations needed for AI models. In contrast, Stanford’s Natural Language Processing Group has 68 GPUs for all of its work.

By Naomi Nix, Cat Zakrzewski and Gerrit De Vynck

March 10, 2024 at 7:00 a.m. EDT

Fei-Fei Li, the “godmother of artificial intelligence,” delivered an urgent plea to President Biden in the glittering ballroom of San Francisco’s Fairmont Hotel in June.

The Stanford professor asked Biden to fund a national warehouse of computing power and data sets — part of a “moonshot investment” allowing the country’s top AI researchers to keep up with tech giants.

She elevated the ask Thursday at Biden’s State of the Union address, which Li attended as a guest of Rep. Anna G. Eshoo (D-Calif.) to promote a bill to fund a national AI repository.

Li is at the forefront of a growing chorus of academics, policymakers and former employees who argue that the sky-high cost of working with AI models is boxing researchers out of the field, compromising independent study of the burgeoning technology.

As companies like Meta, Google and Microsoft funnel billions of dollars into AI, a massive resources gap is building with even the country’s richest universities. Meta aims to procure 350,000 of the specialized computer chips — called GPUs — that are essential to run the gargantuan calculations needed for AI models. In contrast, Stanford’s Natural Language Processing Group has 68 GPUs for all of its work.

To obtain the expensive computing power and data required to research AI systems, scholars frequently partner with tech employees. Meanwhile, tech firms’ eye-popping salaries are draining academia of star talent.

Big tech companies now dominate breakthroughs in the field. In 2022, the tech industry created 32 significant machine learning models, while academics produced three, a significant reversal from 2014, when the majority of AI breakthroughs originated in universities, according to a Stanford report.

Researchers say this lopsided power dynamic is shaping the field in subtle ways, pushing AI scholars to tailor their research for commercial use. Last month, Meta CEO Mark Zuckerberg announced that the company’s independent AI research lab would move closer to its product team, ensuring “some level of alignment” between the groups, he said.

“The public sector is now significantly lagging in resources and talent compared to that of industry,” said Li, a former Google employee and the co-director of the Stanford Institute for Human-Centered AI. “This will have profound consequences because industry is focused on developing technology that is profit-driven, whereas public-sector AI goals are focused on creating public goods.”

Some are pushing for new sources of funding. Li has been making the rounds in Washington, huddling with White House Office of Science and Technology Policy Director Arati Prabhakar, dining with the political press at a swanky seafood and steak restaurant and visiting Capitol Hill for meetings with lawmakers working on AI, including Sens. Martin Heinrich (D-N.M.), Mike Rounds (R-S.D.) and Todd Young (R-Ind.).

Large tech companies have contributed computing resources to the National AI Research Resource, the national warehouse project, including a $20 million donation in computing credits from Microsoft.

“We have long embraced the importance of sharing knowledge and compute resources with our colleagues within academia,” Microsoft Chief Scientific Officer Eric Horvitz said in a statement.

Policymakers are taking some steps to address the funding gaps. Last year, the National Science Foundation announced a $140 million investment to launch seven university-led National AI Research Institutes to examine how AI could mitigate the effects of climate change and improve education, among other topics.

Eshoo said she hopes to pass the Create AI Act, which has bipartisan backing in the House and the Senate, by the end of the year, when she is scheduled to retire. The legislation “essentially democratizes AI,” Eshoo said.

But scholars say this infusion may not come quickly enough.

As Silicon Valley races to build chatbots and image generators, it is drawing would-be computer science professors with high salaries and the chance to work on interesting AI problems. Nearly 70 percent of people with PhDs in AI end up in private industry compared with 21 percent of graduates two decades ago, according to a 2023 report.

Big Tech’s AI boom has pushed the salaries for the best researchers to new heights. Median compensation packages for AI research scientists at Meta climbed from $256,000 in 2020 to $335,250 in 2023, according to, a salary-tracking website. True stars can attract even more cash: AI engineers with a PhD and several years of experience building AI models can command compensation as high as $20 million over four years, said Ali Ghodsi, who as CEO of AI start-up Databricks is regularly competing to hire AI talent.

“The compensation is through the roof. It’s ridiculous,” he said. “It’s not an uncommon number to hear, roughly.”

University academics often have little choice but to work with industry researchers, with the company footing the bill for computing power and offering data. Nearly 40 percent of papers presented at leading AI conferences in 2020 had at least one tech employee author, according to the 2023 report. And industry grants often fund PhD students to perform research, said Mohamed Abdalla, a scientist at the Canada-based Institute for Better Health at Trillium Health Partners and incoming assistant professor at the University of Alberta, who has conducted research on the effect of industry on academics’ AI research.

“It was like a running joke that, like, everyone is getting hired by them,” Abdalla said. “And the people that were remaining, they were funded by them — so, in a way, hired by them.”

Google believes private companies and universities should work together to develop the science behind AI, said Jane Park, a spokesperson for the company. Google still routinely publishes its research publicly to benefit the broader AI community, Park said.

David Harris, a former research manager for Meta’s responsible AI team, said corporate labs may not censor the outcome of research but may influence which projects get tackled.

“Anytime you see a mix of authors who are employed by a company and authors who work at a university, you should really scrutinize the motives of the company for contributing to that work,” said Harris, who is now a chancellor’s public scholar at the University of California at Berkeley. “We used to look at people employed in academia to be neutral scholars, motivated only by the pursuit of truth and the interest of society.”

Tech giants procure huge amounts of computing power through data centers and have access to GPUs. These resources are expensive: A recent report from Stanford University researchers estimated that Google DeepMind’s large language model, Chinchilla, cost $2.1 million to develop. More than 100 top artificial intelligence researchers on Tuesday urged generative AI companies to offer a legal and technical safe harbor to researchers so they can scrutinize their products without the fear that internet platforms will suspend their accounts or threaten legal action.

The necessity for advanced computing power is likely to only grow as AI scientists crunch more data to improve the performance of their models, said Neil Thompson, director of the FutureTech research project at MIT’s Computer Science and Artificial Intelligence Laboratory, which studies progress in computing.

“To keep getting better, [what] you expect to need is more and more money, more and more computers, more and more data,” Thompson said. “What that’s going to mean is that people who do not have as much compute [and] who do not have as many resources are going to stop being able to participate.”

Tech companies like Meta and Google have historically run their AI research labs to resemble universities where scientists decide what projects to pursue to advance the state of research, according to people familiar with the matter who spoke on the condition of anonymity to discuss private company matters.

Those workers were largely isolated from teams focused on building products or generating revenue, the people said. They were judged on influential papers they published or notable breakthroughs — similar to metrics used for their university peers, the people said. Top AI Meta scientists Yann LeCun and Joelle Pineau hold dual appointments at New York University and McGill University, blurring the lines between industry and academia.

In an increasingly competitive market for generative AI products, research freedom inside companies could wane. In April, Google announced it was merging two of its AI research groups — DeepMind, which it acquired in 2014, and the Brain team from Google Research — into one department called Google DeepMind. Last year, Google started to take more advantage of its own AI discoveries, sharing research papers only after the lab work had been turned into products, The Washington Post has reported.

Meta has also reshuffled its research teams. In 2022, the company placed its Fundamental AI Research team, known as FAIR, under the helm of its VR division Reality Labs and last year reassigned some of the group’s researchers to a new generative AI product team. Last month, Zuckerberg told investors that FAIR would work “closer together” with the generative AI product team, arguing that while the two groups would still conduct research on “different time horizons,” it was helpful to the company “to have some level of alignment” between them.

“In a lot of tech companies right now, they hired research scientists that knew something about AI and maybe set certain expectations about how much freedom they would have to set their own schedule and set their own research agenda,” Harris said. “That’s changing, especially for the companies that are moving frantically right now to ship these products.”


Go to Top