The Nuclear Regulatory Commission has proposed changing how it writes and enforces parts of its radiation-exposure rules, a smaller move than nuclear industry boosters had been waiting for. The agency would remove references to ALARA, the long-used standard meaning “as low as reasonably achievable,” and replace it with a more explicit, tiered approach for controlling doses below legal limits.
The practical effect looks modest. In its proposal, the NRC estimates the changes would save regulated industries about $9.5 million a year across nuclear power, medical uses, research, and other licensed activities. Spread across the 57 nuclear power plants in the United States, if all of that savings were assigned to power plants alone, it would average a little over $150,000 per plant. That is rounding dust in nuclear construction economics, which is the part of the debate that keeps getting the microphones.
The proposal lands after President Donald Trump ordered a reconsideration of nuclear regulation as part of his administration’s push to restart nuclear plant construction in the US. His executive order criticized safety models that assume no fully safe level of radiation exposure, saying they lack a sound scientific basis and produce irrational outcomes. The NRC is declining to go that far.
The model stays
The key scientific model at issue is called linear no-threshold, or LNT. In plain English, it treats radiation risk as rising in proportion to dose and does not assume a dose so low that biological risk disappears. The NRC says that model remains appropriate for radiation-protection rules because there is no “consensus-supported, regulation-ready alternative” available.
The agency also says a true no-effect threshold is unlikely unless DNA repair worked perfectly at low doses or a single radiation track could not cause a biological effect. That is a dry way of saying the cells are not running enterprise-grade error correction, and regulators are not ready to pretend otherwise.
The NRC has been asked before to move toward hormesis, the idea that very low radiation doses may trigger protective cellular repair. During Trump’s first term, the agency rejected petitions asking it to adopt that approach. In the new proposal, it again leaves LNT standing.
ALARA gets the hook, mostly
ALARA followed logically from LNT: if any radiation dose carries some risk, exposures should be pushed down where reasonably possible. The messy word has been “reasonably.” Critics have argued that the standard encouraged endless pressure to reduce dose whenever any reduction was technically available, even when the gain was tiny or the cost was high.
The NRC now acknowledges that implementation has been uneven. The agency says ALARA has suffered from unclear stopping points, subjective judgments, and the potential for selective or inconsistent enforcement. The proposed rule would replace the label with a graded system that sets exposure thresholds below levels where evidence shows radiation effects would be expected, then requires stronger dose-reduction measures as exposures rise.
That is still optimization of radiation protection, and the proposal’s language appears to tug in two directions. The NRC is keeping a model that says risk does not vanish at low doses while regulating through thresholds below formal limits. It also describes its new approach using concepts that overlap with the ALARA framework it wants to stop naming. Regulatory drafting: a machine for turning a label change into a PDF.
The proposal also updates rules for exposure-monitoring equipment, reflecting newer technology since the NRC last revised those requirements.
For current licensees, the agency says compliance status would carry over. Facilities that already meet the rules would not need to change practices unless they decide the new structure saves money. That makes the proposal a cleanup and clarification exercise more than a deregulatory detonation. The word ALARA may be headed out of the rulebook, but the risk model underneath it is staying put.
This story draws on original reporting from Ars Technica.
