Nuclear Plants Reported More Than 1,100 Drug or Alcohol Testing Violations Last Year. The NRC Now Wants to Scale Back Testing.

Nuclear plant workers are still going to work and failing drug and alcohol tests. Others are caught trying to cheat.

The most recent numbers: in 2025, companies that hold licenses to operate nuclear power plants in the U.S. reported that 1,174 workers failed these tests, the highest total in five years, according to a slide presentation by NRC officials earlier this year.

And a record 293 nuclear workers were caught trying to cheat the tests.

Now the NRC is proposing a change to regulations that would cut by half random testing for most licensee employees (employees who work directly on the payroll of power companies that operate the plants), broaden the credentials accepted for some fitness-for-duty professionals, reduce certain testing and training requirements, and end licensee audits of federally certified testing laboratories.

The proposal follows President Trump’s Executive Order 14300, which directed the NRC to accelerate reactor licensing and undertake a wholesale revision of its regulations. The order calls for less regulatory burden and faster decisions on reactor applications.

The changes have drawn criticism from the administration of New York Gov. Kathy Hochul—herself generally a supporter of nuclear power—and the International Brotherhood of Electrical Workers.

“In New York’s view, NRC has not provided a substantive analysis to support its proposal to reduce the annual testing rate of most licensee employees from 50 percent to 25 percent,” wrote the state’s NRC liaison officer–designee to the agency as part of its public comment process. “… Before cutting the amount of random testing in half, NRC should conduct a comprehensive analysis of the potential impacts of such a change.”

DOWNLOAD: The NRC’s PowerPoint on proposed changes to its fitness-for-duty testing.

Proposed changes to testing are among a list of many other changes to NRC policies under EO 14300, and come at a time when the industry is racing to bring new nuclear power sources online to feed a growing demand—led by planned, proposed and future hyperscale data centers.

Consider that, in Iowa, Google has contracted to bring back the decommissioned nuclear power plant to supply power to a grid for its own data center plans. And Microsoft has entered into a power purchase agreement that will reopen the site of the former Three Mile Island nuclear power plant in Pennsylvania. That site is well-known for the nuclear accident in 1979 that led to releases of radiation and significant damage to the core of one of two nuclear reactors at the site, which sits on the Susquehanna River outside Middletown, Pa.

(Microsoft’s contract is on the other reactor that shut down in 2019.)

Following the 1979 Three Mile Island accident, the NRC enacted sweeping changes to its regulatory regime and followed up several times after that. EO 14300 seeks to roll back some of those regulations including drug- and alcohol testing.

What Random Testing Is Finding

On July 17, 2025, a licensed reactor operator at the Watts Bar Nuclear Plant in Tennessee was pulled for a random drug test during a training week. The operator was a Tennessee Valley Authority employee.

Records show that two days later, before the result came back, he performed licensed reactor duties. When the lab reported on July 28, 2025, the operator’s specimen contained 248 nanograms per milliliter of cocaine metabolites — well above the federal cutoff.

Two months later, when he tested to regain access to the plant, he came back positive again.

This reactor operator’s case shows what unannounced testing can detect among established nuclear employees when no supervisor has identified a reason to order a test.

The NRC would continue random testing for licensed operators at the current rate. But it wants to cut the rate in half for most other employees of nuclear-plant licensees, eliminating an estimated 12,600 tests a year as the federal government presses the agency to accelerate reactor licensing and rewrite its regulations.

The changes sit inside a rule called “Modernizing Security Requirements,” published June 26, 2026 to carry out EO 14300, President Trump’s order to reduce regulatory burdens across the NRC. Included in a much broader rulemaking dominated by physical security is a substantial rewrite of 10 CFR Part 26 — the rules that require licensees to test workers and manage fatigue.

The NRC would cut the annual random-testing rate from 50 percent to 25 percent for most licensee employees—specifically, those outside the safety- and security-significant groups that would retain the current rate. The agency says “35 years of operating experience demonstrates low positive rates in this worker population.”

The reduction would not apply to contractors and vendors, licensed reactor operators, security personnel, fitness-for-duty staff or certain supervisors directing safety- and security-related work. Those groups would remain subject to the 50 percent rate. Pre-access, for-cause and follow-up testing also would remain in place.

But the rule change goes well beyond one testing rate.

It would end annual licensee audits of the HHS-certified laboratories that analyze the specimens. The NRC says those audits are redundant because the federal laboratory-certification program independently inspects the laboratories twice a year and the nuclear program would retain blind-sample performance testing. It would eliminate annual fatigue-management reports.

It would stretch refresher training from every year to every two years, permit video determinations of fitness in for-cause cases and double the medical-evaluation window for a “shy bladder” case from five business days to ten.

The rule would also broaden who can serve as a substance-abuse expert or medical review officer, let licensees escort some short-duration construction workers instead of placing them in a fitness-for-duty program, and allow alternative fatigue controls for as long as 60 days during a “sequestration” event such as severe weather or a public-health emergency.

The NRC calls much of this housekeeping: align Part 26 with Department of Transportation practice and Department of Health and Human Services guidance, remove duplicate paperwork, and let licensees run leaner programs as the country tries to build reactors faster.

The NRC’s case

Nuclear power is not the only safety-sensitive industry that subjects employees to drug and alcohol testing. Under federal aviation rules, airlines must randomly test at least 25 percent of covered workers for drugs and 10 percent for alcohol each year; the testing pool includes pilots, flight attendants, mechanics and dispatchers. 

The aviation program also requires testing before employment and in certain post-accident, reasonable-suspicion, return-to-duty and follow-up situations.

For its part, the NRC says it is narrowing requirements according to risk, not weakening the fitness-for-duty program.

Random positive rates among licensee employees—a broader category that includes the workers covered by the proposed reduction—have held near 0.3 percent for years, far below contractor rates.

The agency also says fatigue data will remain available to inspectors even if licensees no longer file an annual report.

Some changes could strengthen detection. The rule would expand oral-fluid testing to all testing conditions, allowing directly observed collections that are harder to subvert. And the 2022 decision to lower drug-testing cutoffs — the change now finding more cocaine — is evidence that the NRC can tighten the program when the data support it.

The comments were not uniformly in favor of more testing. Workplace drug-testing consultant Nadine Wentzell urged the NRC to end blind-specimen laboratory tests, which the Department of Transportation abandoned in 2017 as redundant. Industry commenters asked the agency to go further in other directions, including testing for impairing substances such as kratom and allowing rapid point-of-collection screens.

This is not a simple fight between testing and no testing. It is a fight over where the line should be drawn, and what evidence the NRC should have to move it.

The NRC’s own numbers

The timing is hard to miss. On July 16, 2026 — three weeks after the rule was published and 11 days before comments closed — an NRC fitness-for-duty specialist presented the program’s 2025 results to an industry workshop in Knoxville.

The slides were still marked draft. The direction of the numbers was not.

Licensees tested 107,977 individuals in 2025 and reported 1,174 who tested positive or were otherwise noncompliant. The total includes positive drug and alcohol tests, adulterated or substituted specimens and refusals to test.

That was the highest such total in at least five years, up from 1,089 in 2024 and 1,112 in 2021, even as the number of reporting facilities fell from 69 to 60. Marijuana accounted for roughly half of the positive results. Alcohol was next.

The aggregate figure combines employees and contractors, different testing conditions and violations ranging from confirmed drug use to refusals and substituted specimens. It shows that the fitness-for-duty program remains active and consequential. By itself, it does not establish the positive rate among the specific licensee employees whose random-testing rate would be reduced.

Two trends jump out.

The first is cocaine. After the NRC lowered testing cutoffs in 2022, the cocaine positive rate rose. The cocaine-positive rate cited in the presentation went from 0.079 percent in 2023 to 0.115 percent in 2024, then eased to 0.103 percent in 2025 — still above the 2023 level.

The second: people at the controls.

Licensed reactor operators would not be included in the proposed testing reduction. Their increase nevertheless demonstrates the potential consequences of drug use within the workforce and the continuing value of unannounced testing.

The NRC recorded 12 reportable fitness-for-duty violations by licensed reactor operators — the NRC-licensed personnel authorized to operate a reactor’s controls — in 2025. That was twice the total in each of the previous three years and the most in a series going back to 2017. Seven of those 12 cases were found through random testing.

By June 25, 2026, five more licensed-operator violations had been reported in 2026. Four were caught by random testing.

The NRC’s presentation said it plainly: “Licensed reactor operator positives — significant increase in 2025, continuing into 2026.”

The reactor operator case

According to the NRC’s March 26, 2026 Notice of Violation, TVA reported the operator’s first confirmed positive in late July 2025 and a second in September. The July sample came from a random test and contained 248 ng/ml of cocaine metabolites, above the 150 ng/ml cutoff in 10 CFR 26.163. The September test, required before he could regain unescorted access, came back at 197 ng/ml.

The NRC found that he had performed licensed activities “under the influence of cocaine, an illegal substance,” in violation of 10 CFR 55.53(j). It classified the case as Severity Level III, among the more serious levels on its enforcement scale.

The worker wrote that the cause was unknown and “probably due to environmental reasons.” The agency rejected that explanation outright.

“Environmental exposure to cocaine or its metabolites is not known to produce confirmed positive FFD test results,” the NRC wrote. The Part 26 cutoffs are designed to prevent false positives from incidental or passive exposure. Without a legitimate explanation, the agency said, a confirmed positive is treated as actual use.

TVA had already asked the NRC to terminate his operator license.

The 2026 case files

The Watts Bar operator’s case produced a public enforcement letter. It was not an isolated event.

The single-positive-test reports filed with the NRC in the first half of 2026 show the testing program working across the fleet. The reports do not name workers, but they identify the plant, the date, the job, the substance and what happened.

A sample:

Palo Verde Generating Station, Arizona, March 18, 2026 — cocaine, caught at random. A licensee employee in radiation protection tested positive on a random screen. The public report does not provide enough information to determine whether the employee belongs to the group whose random-testing rate would be reduced.

Alvin W. Vogtle Generating Plant, Georgia, January 14, 2026 — cocaine and a cheating device. A licensee security employee was tested for cause. A collector found a subversion device in the worker’s bag. The worker said he planned to use it if called for a random test. The observed sample was positive for cocaine.

Vogtle, May 21, 2026 — alcohol, already on the clock. A contractor tested positive on a random test at 0.03 BAC after being in work status for at least an hour.

The proposal would not reduce testing for the Vogtle security employee or contractor. The public record does not establish whether the Palo Verde employee belongs to the affected group, but the three cases show the program working under different conditions.

Most of those testing requirements would remain unchanged. The narrower question is whether the NRC has shown that other licensee employees can safely be screened half as often.

Cheating is rising

The other trend in the NRC’s data is subversion — the agency’s term for cheating on a test by adulterating or substituting a specimen, or by refusing once selected.

The NRC counted 293 cheating attempts in 2025, up from 232 in 2024 and the most in five years. Subversion made up 29 percent of all drug-testing violations. Contractor and vendor workers accounted for 94.5 percent of the attempts. Eighty-two percent of sites reported at least one.

The split between licensee employees and contractors runs through the data. Contractors and vendors test positive at two to three times the rate of licensee employees. Their pre-access positive rate reached 1.62 percent in 2025, the highest since 2003. The rate for licensee employees was 0.69 percent.

Invalid specimens — every one classified as subversion — increased for a third consecutive year, reaching 31 in 2025.

Pre-access testing catches the largest share of violations, usually 66 to 75 percent and 74.5 percent in 2025. The proposed rule leaves that screening intact.

Random testing catches much of what remains. The rule would preserve the current rate for operators and contractors — the workers responsible for most subversion attempts — but reduce it for most other licensee employees.

The cheating data therefore do not directly refute the proposed population split. They do show that workers continue trying to defeat testing at a time when the NRC is reducing several forms of oversight.

The case against the cut

The State of New York made the missing evidence the center of its objection. New York is home to four operating nuclear reactors and has committed to building at least five gigawatts of new nuclear capacity. It is not arguing against the industry’s expansion.

Through its state energy agency, NYSERDA, the Hochul administration took aim at other aspects of the proposed drug policy changes.

New York also urged the NRC to keep annual fatigue-management reports. The state argued that the agency’s attention to those reports may have helped drive the decline in waivers the NRC now cites as a reason to stop collecting them.

The International Brotherhood of Electrical Workers, representing more than 10,000 workers at more than 55 nuclear facilities, tied the changes to the industry’s history.

The union said several provisions would weaken programs created after Three Mile Island and could “increase the risk of accidents and jeopardize the nuclear renaissance.” On video determinations of fitness, it asked for safeguards against arbitrary testing and surveillance.

A citizen comment from a man identified as Fred Schofer, in a detailed technical comment, made the same point from another direction: the data the NRC says it relied on are not in the public record.

The agency cites “approximately 35 years of FFD program performance data,” he wrote, but identifies no NUREG (a report in the NRC’s numbered technical-report series), information notice or annual aggregate report that documents the comparison. Nor does it provide the absolute positive rates, sample sizes or confidence intervals. He asked the NRC to put the dataset, an aggregate report or a citable summary in the docket before acting.

An anonymous commenter pointed back to the NRC’s own Knoxville presentation. The number of HHS-certified laboratories fell from 30 in 2016 to 14 by mid-2026, the commenter wrote, while standard test panels remain too narrow to capture newer substances. The comment urged the agency to add hair and oral-fluid testing instead of loosening the program.

Jacquelyn Drechsler, a social worker who lives near a plant, called the end of annual fatigue reporting “a disservice to all who have responsibility for the safety of the plant.” Her one-word summary was: “shameful.”

What happens next

The comment period for these proposed changes is closed. NRC staff will review the record and prepare a final rule. There is no statutory deadline, and the Commission can keep, change or drop any part of the proposal. The NRC’s public rulemaking schedule currently identifies January 12, 2027, as its target date for a final rule.

New York asked for the most direct answer to the dispute: keep the current testing rate, or publish the analysis behind cutting it, before the rule becomes final.

The United States is trying to expand nuclear power for the first time in a generation. That means more contractors, more construction crews and more first-time entrants to an industry where one impaired operator can become a national story.

The NRC says 35 years of data support testing most other licensee employees half as often. Commenters say the agency has not placed the underlying analysis in the public record. Before making the change final, the agency should show its work.