Sitting above a Pacific subduction fault in Shizuoka, a reactor’s safety data hid a pool of nearly 30,000 seismic readings and reported only 20, and 208 of 225 assessments were later found to contain improper practices
The Pacific ocean is visible from the control room windows at Hamaoka, a nuclear station planted on a narrow strip of Shizuoka coastline directly above one of the world’s most feared fault systems.
For more than a decade, engineers there were working toward a restart.
The numbers they submitted to Japan’s nuclear regulator, it turned out, were not the numbers that a proper calculation would have produced.
The company announced the resignations of its president and chairman, as well as the withdrawal of its application for safety review to restart Hamaoka’s No.
An external tip reached regulators in early 2025, and what the investigation uncovered over the following months was not a rounding error or a clerical slip.
It was a systematic choice, made across departments, to present a far smaller earthquake as the one the plant needed to survive.
How a figure drawn from nearly 30,000 data points became a figure drawn from 20 is the question that pulled two reactors off the restart list entirely.
How cherry-picking a handful of waves out of tens of thousands kept the danger number low
Seismic safety screening for a nuclear reactor requires an operator to estimate the maximum ground motion the site could experience, a figure that then sets the structural and mechanical standards everything else is built around. A lower maximum means lower engineering costs, a shorter construction timetable and a smoother path through a regulatory review that can run for years.
The utility told Japan’s Nuclear Regulation Authority it had generated 20 sets of seismic waves under different calculation conditions and selected the one closest to the average as the representative wave. In reality, the company had asked a contractor to select data favorable to the company from more than 1,000 data points. In the most extreme cases, the cherry-picking involved nearly 30,000 data points.
So the gap between what regulators were told and what actually happened was not a matter of statistical nuance. The company allegedly underestimated the maximum seismic ground motion by selecting favorable data while suggesting that the figures came from proper calculation. The result was a safety envelope drawn around a smaller earthquake than the ground beneath Hamaoka may actually produce.
The place that made the numbers matter more than anywhere else
Hamaoka is not a generic coastal site. The plant is located in an area where a Nankai Trough megaquake in the Pacific is expected to occur, and all of its reactors were shut down at the Japanese government’s request following the 2011 Fukushima disaster, which was triggered by a massive earthquake and tsunami. That history hung over every data point in the screening files.
The utility applied to the regulator between 2014 and 2015 for screening to determine whether the No. 3 and No. 4 reactors at Hamaoka complied with new, stricter safety rules, aiming to restart the two reactors. The review ran for years, and in 2023 the regulator broadly approved the company’s assumption for the maximum seismic ground motion. That approval now looks like the moment the manipulated figures did their work.
Because the plant sits above a subduction zone fault, regulators and the public had reason to expect the seismic bar to be set higher than at almost any other nuclear site in Japan, not quietly lowered through contractor-selected waveforms.
A cover-up inside the cover-up: 69 data sets rewritten after hearings began
The NRA became aware of the data fraud through an external tipoff in February 2025. Once hearings began, something worse emerged. The NRA secretariat said the company retroactively collected and rewrote data linked to the cherry-picking after it began hearings over the scandal in May of the previous year, affecting 69 of the 225 sets of data being investigated.
That sequence, a data-manipulation scheme followed by an evidence-rewriting campaign while regulators were actively looking, is what pushed the NRA chairman to use language rarely heard from a nuclear watchdog. “We suspect there was a cover-up,” NRA Chairman Shinsuke Yamanaka told a news conference. He added separately that he felt there was malicious intent behind what had been done.
A third-party committee of external lawyers completed the formal tally. The committee found that at least 208 of the 225 seismic ground motion assessments suspected of containing problems involved improper practices, including false explanations and concealment. Of 225 files, just 17 were clean.
What a lower seismic number actually costs, and who paid it in the end
A lower maximum seismic ground motion could reduce earthquake-proofing costs for plant facilities and shorten construction periods. That is the institutional logic behind a calculation that spanned multiple departments and, investigators found, may have reached senior management. The economic pressure to restart Hamaoka was real: the plant is the only nuclear power station operated by the utility, meaning years of idled capacity sat entirely on this one site.
The eventual cost of the manipulation was far larger than any engineering savings. The company announced the resignations of its president and chairman, as well as the withdrawal of its application for safety review to restart Hamaoka’s No. 3 and No. 4 reactors. Both executives stepped down on September 30, leaving the company to rebuild trust under new leadership. Japan’s regulator simultaneously began drafting a law revision that would make falsification in reactor safety applications subject to criminal penalties, a provision that did not exist when the manipulation was carried out. You can read about another layer of the nuclear regulatory apparatus here, where the NRC issued the first US construction permit for a small modular reactor after a full safety review process of the kind Hamaoka’s operators sought to shortcut.
What the scandal leaves open, and whether the fix can outlast the lesson
The scandal has rekindled public anxiety and distrust over nuclear power, in a setback to the government’s energy policy at precisely the moment Japan was counting on a nuclear revival to meet rising electricity demand driven by data centers and decarbonization commitments. Hamaoka’s two idle units will now remain offline indefinitely, and the restart applications must be resubmitted from scratch, if they are ever resubmitted at all.
Japan’s Nuclear Regulation Authority decided to consider a law revision aimed at making any falsification in nuclear reactor safety screening applications subject to criminal penalties. The secretariat is investigating the utility to determine whether multiple departments and senior management were involved, as well as the motives behind the misconduct. That investigation had not concluded as of publication.
The harder question is structural. A safety screening process that ran for nearly a decade, involved hundreds of data submissions and was formally approved in 2023 still produced the wrong answer. The regulator caught it only because an anonymous external source sent a tip. Investigators have not confirmed who filed it or why it took until 2025 to surface. For the communities on the Shizuoka coast, the unresolved irony is that the seismic danger the data was crafted to minimize remains entirely unchanged. The fault beneath Hamaoka has not moved. Only the paperwork above it has. The broader challenge of aligning institutional incentives with genuine safety standards is one the industry is watching closely, as new reactor designs and licensing frameworks explored in projects like advanced reactor assessments in the UK try to build independent verification into earlier stages of the process.
Hugo is an engineer with strong technical expertise. Multilingual from an early age, his writing combines technical clarity with a strong interest in science and energy.
