Docket HHS-OASH-2026-0397 • Questions 2–18
Prepared for review • October 6, 2026
John Coates, Founder, RF Safe
Comment-box cover note
Please see the attached RF Safe response to questions 2–18 of HHS-OASH-2026-0397. It distinguishes peer-reviewed findings, personal motivation, and proposed research. We request sustained independent research on developmental timing, waveform dependence, susceptibility, and persistent regulatory responses. Our priority pilot asks one simple question: after an altered planarian head has formed, does a defined electromagnetic exposure change how soon it reverts to species-typical morphology? The pilot is proposed research, not evidence that everyday exposure causes harm. References and active source URLs appear in the attachment.
Question 2 — Perspective and source types
I submit this response as founder of RF Safe, established in 1998, and as an engineering and public-health advocate with a longstanding interest in electromagnetic exposure. RF Safe also has commercial interests in exposure-reduction accessories and electromagnetic technologies; those interests should be visible when our recommendations are evaluated. We are not presenting this comment as an independent clinical trial or a systematic review.
Our response draws on government toxicology reports, peer-reviewed studies, official legal and regulatory documents, and explicitly identified conceptual proposals. The Cellular Latent Learning Model, ceLLM, is our parent theory of cellular interpretation and persistent response state. S4–Mito–Spin is its proposed mechanism layer, incorporating voltage-sensitive channels, mitochondrial/redox amplification, and spin-sensitive chemistry. Neither the complete theory nor the proposed planarian exposure assay is established by the cited literature.
Question 3 — Personal motivation and reported experience
RF Safe’s founding is inseparable from a family loss. My firstborn daughter died of anencephaly in 1995. I do not claim that electromagnetic exposure was proven to cause her condition, and I am not submitting a documented exposure–disease attribution. The loss is the reason developmental timing has remained central to my work.
In 1997, Farrell and colleagues published a morphology study involving more than 2,500 White Leghorn chick embryos across five experimental campaigns. They reported increased abnormalities under selected weak pulsed and sinusoidal magnetic-field conditions, while acknowledging differing results across the wider literature and variability among campaigns. A companion study using a 4-µT, 60-Hz field reported stage-dependent changes in ornithine decarboxylase activity and an approximately threefold abnormality rate, with most malformations involving neural tubes. These experiments did not diagnose my daughter or establish a human prenatal risk estimate. They were important in directing me into this research literature. [1,2]
I also experienced childhood cancer requiring removal of a kidney while growing up near military radar infrastructure. I do not have contemporaneous dosimetry establishing causation. These experiences explain my interest; the scientific assessment must rest on the studies and their exposure conditions. I have kept this account limited rather than providing identifiable family medical records.
Question 4 — Research observations: tissue selectivity, density gating, and ceLLM
We offer observations from published research and a framework for testing their biological interpretation. NTP Technical Report 595 reported clear evidence of carcinogenic activity in male rats based on malignant heart schwannomas under its studied GSM-modulated RF conditions, and some evidence based on malignant brain gliomas. Ramazzini’s lifetime 1.8-GHz study reported a statistically significant increase in heart schwannomas in males at the highest exposure. Its increase in malignant glial tumors in females was not statistically significant. The convergence is clearest for cardiac schwannomas; the brain results are not identical replications. The WHO-commissioned 2025 animal-cancer systematic review rated evidence for selected glioma and heart-schwannoma outcomes as high certainty. These findings warrant a mechanistic research program, while numerical human risk remains a separate question. [4,5,12]
Density gating: why the responding cell matters
RF Safe proposes that susceptibility depends partly on the abundance and functional state of field-coupled transducers, mitochondrial/redox amplification, and the capacity to restore signaling. This is the density-gating component of ceLLM’s S4–Mito–Spin mechanism layer. Its central question is why a defined exposure produces a response in one cell population but little measurable response in another. Absorbed dose remains necessary, but may not be sufficient to predict a biological response if coupling and recovery differ among cells. [20]
In this proposal, S4 denotes the charged voltage-sensing segment of relevant voltage-gated channels in cellular membranes. A perturbation of channel activity could alter calcium timing; mitochondrial calcium handling could then amplify or buffer the response through metabolism and redox chemistry. Spin-sensitive reactions supply another candidate field-responsive route, potentially interacting with this pathway. These are linked candidate processes, not a claim that every mitochondrial calcium transporter contains an S4 sensor or that all field responses must begin at S4. A calcium change alone cannot identify its initiating transducer. [7–9,20]
The relevant unit is the tumor’s cell of origin and its local environment, not simply the organ containing it. Cardiac schwannomas arise from Schwann cells, not cardiac muscle. The mitochondrial abundance of cardiomyocytes therefore cannot by itself explain those tumors. Schwann cells support and myelinate axons; describing them as the cells that conduct action potentials or asserting that they are among the most channel-dense cells requires evidence not supplied by these tumor studies. Density gating needs measurements of the proposed transducers, mitochondrial function, and recovery in the implicated Schwann and glial populations, alongside appropriate comparison cells.
Null skin findings are part of the test
Jyoti and colleagues’ 2025 human skin-cell study found no meaningful exposure-associated gene-expression or DNA-methylation changes under its tested 5G conditions. Other skin-cell research and the SEAWave project also provide important comparisons. These results are compatible with a conditional susceptibility model, but do not establish that lower S4 or mitochondrial density caused the nulls. Skin includes multiple cell types and physiological states. Differences in waveform, dose, duration, culture conditions, and endpoint sensitivity must be considered before comparing short cell experiments with lifetime animal tumor studies. [21–23]
The strong density-gating prediction is prospective: under a specified, comparable exposure, independently measured transducer activity, mitochondrial/redox amplification, and recovery capacity should predict the ordering of cellular responses. If these properties fail to predict held-out results, or simpler dose and stress models predict them equally well, the proposed explanation loses support. A null result cannot automatically be rescued by asserting an unmeasured threshold. This requirement turns a possible explanation of existing findings into a testable research program.
Cord-blood cells: response without persistent damage
Durdik and colleagues’ 2019 laboratory study exposed human umbilical cord-blood cells, including CD34+ stem/progenitor cells, to pulsed mobile-phone signals. It found a transient ROS increase after one hour of UMTS exposure that was no longer evident three hours afterward, without a persistent difference in DNA damage, preleukemic fusion genes, or apoptosis. ROS response also varied with cellular differentiation. This is especially relevant to a state-dependent framework: measurable interaction, successful recovery, and lasting injury are different outcomes. The study does not establish S4 or mitochondrial density as the cause of the differentiation-dependent response. [24]
Susceptibility and calcium coding
Sousouri and colleagues’ 2025 sham-controlled study of 34 volunteers reported a genotype-dependent sleep-spindle response to 3.6-GHz exposure in CACNA1C rs7304986 T/C carriers. Kim and colleagues’ 2026 engineered gene-switch study identified Cyb5b as an essential mediator and likely sensor, with activation depending on rhythmic calcium dynamics rather than generic influx. Specified spin-chemistry systems also show field-responsive reaction behavior. Together these studies motivate measuring receiver state and signal timing, rather than assuming bulk calcium or average power captures every relevant response. They do not establish a shared causal pathway, a disease diagnosis, or equivalent effects from household exposures. [6–9]
Low fidelity and the proposed meta-disease state
ceLLM is RF Safe’s parent theory of cellular interpretation and persistent regulatory state; S4–Mito–Spin supplies candidate coupling and amplification mechanisms. “Low fidelity” means a hypothesized loss of accuracy or timing in a defined biological task. “Meta-disease” denotes a proposed vulnerability state in which altered regulation changes responses to other stressors, rather than one specific disease caused by RF alone. These terms need operational measurements; the cell-as-language-model analogy is conceptual, not demonstrated cellular machinery.
The proposed chain is conditional: a defined exposure perturbs a transducer; calcium or redox signaling changes; restoration is incomplete or repeated perturbation changes regulatory state; a later challenge then elicits an altered response. Infection, nutritional status, metabolic stress, genotype, and other exposures may modify each step. Under this hypothesis, the same input could be tolerated in one state but contribute to dysfunction in another. A transient ROS response with recovery is evidence against assuming every interaction necessarily enters a chronic harmful state.
Immune tolerance, inflammatory resolution, and metabolic regulation are therefore research targets, not diseases already proven to arise through ceLLM. Established inflammatory-memory studies show that prior challenges can alter later responses in epithelial stem cells and brain innate immune cells. They provide precedent for persistent biological state, not proof that wireless exposure produces the same memory. An RF-to-memory-to-dysfunction pathway must be demonstrated separately. [27,28]
The framework’s value is that it connects tissue selection, acute signaling, recovery, and exposure history into predictions HHS can fund. Its explanatory advantage over alternatives remains to be tested. We request independent replication, measurements in relevant cell populations, exposure–washout–challenge experiments, and publication of negative findings. The planarian pilot in Question 16 asks a narrower question about reversion timing; a result would guide this program without proving or disproving the complete human immune and metabolic hypothesis.
Question 5 — Standards, classification, and needed changes
Federal health agencies should evaluate FCC exposure rules, ICNIRP’s 2020 RF guidelines, IEEE C95.1-2019, and relevant low-frequency guidance as distinct frameworks with defined frequency ranges, populations, exposure metrics, and protected endpoints. Nerve stimulation, heating, and electromagnetic interference with medical devices must remain differentiated. It is inaccurate to describe every provision as relying exclusively on SAR. [10,11]
The density-gating proposal additionally calls for biological characterization: cell type, differentiation state, relevant channel activity, mitochondrial/redox function, and recovery capacity. These are research variables to test alongside physical exposure, rather than validated replacements for compliance metrics.
Exposure classification should retain carrier frequency, electric and magnetic fields, absorbed dose where applicable, peak and average values, duty cycle, burst timing, modulation, duration, geometry, and simultaneous sources. Recording these properties is necessary to test waveform dependence, even where biological relevance remains uncertain. Existing standards do address some short-duration and pulsed exposures; our concern is whether their health endpoints sufficiently cover proposed signaling and persistent-state effects.
The proposed planarian pilot compares defined continuous and pulsed magnetic exposures during reversion. Equal peak amplitude alone does not isolate temporal structure: RMS field, duty cycle, spectrum, and induced electric field can differ. The study must report those differences. A 217-Hz magnetic pulse train is not equivalent to a GSM RF carrier with a 217-Hz envelope.
For exposure minimization, HHS should evaluate reproducible reduction strategies, including distance, wired connections where feasible, and engineering changes, while testing performance and unintended consequences. The research program should examine waveform and susceptibility hypotheses directly rather than infer them from compliance alone.
Question 6 — Evidence relevant to current limits
An exposure limit is a regulatory boundary, not itself an exposure that causes an effect. The substantive question is whether the evidence supports the level and scope of protection it supplies.
NTP used whole-body rat SARs of 1.5, 3, and 6 W/kg, above the 0.08-W/kg general-public whole-body restriction. It should not be represented as a below-public-limit experiment. Its chronic exposure schedule, animal biology, and findings still matter to hazard assessment, but human relevance requires a separate analysis. [4]
HHS should publish an endpoint-specific assessment of the evidence, including study quality, exposure validity, temperature control, replication, developmental timing, and uncertainty. It should explain which outcomes each existing limit addresses and identify remaining research questions without equating every biological change with harm.
Question 7 — Available exposure methods
Available methods include calibrated field probes, personal exposimeters, spectrum and time-domain measurements, source and device logs, propagation models, and computational or experimental SAR assessment. These methods have different strengths: direct field measurement captures a location and time; modeling estimates internal dose; personal monitoring can relate exposure to behavior but has placement, bandwidth, and calibration limitations.
Temporal characterization is technically possible. The problem is inconsistent collection and integration of it in biological and population studies, not the absence of all waveform-measurement methods. RF carriers and actual low-frequency electric or magnetic fields should be measured separately. A low-frequency envelope is not automatically an equivalent independent ELF field inside tissue.
Question 8 — Exposure-assessment improvements
HHS should promote interoperable measurement protocols that preserve average and peak values, waveform information, frequency, source proximity, duration, and uncertainty. Studies should characterize the exposure at the biological sample or participant rather than only quote transmitter settings.
Link longitudinal exposure records to individual activity and relevant covariates, while protecting privacy. Laboratory work should use calibrated measurements, sham conditions, temperature checks, blinded endpoint assessment, and independent replication. Our planarian pilot adds a simple timing endpoint to this toolkit; it does not replace dosimetry or validate a universal measure of signaling fidelity.
Question 9 — Disclosure
Disclose the applicable compliance standard, test configuration, averaging procedure, body separation assumptions, operating modes, and relevant limitations. SAR measures absorbed RF power per tissue mass; it is not a direct reading of a consumer’s tissue temperature or total lifetime exposure.
Provide accessible information on practical exposure reduction and maintain technical data suitable for researchers. For infrastructure, publish measured or modeled exposure ranges, uncertainty, monitoring dates, and source conditions. Waveform information should be available in a usable technical form without presenting unvalidated timing patterns as established hazards.
Question 10 — Surveillance
Disease registries, occupational records, environmental monitoring, longitudinal cohorts, and voluntary reports can supply complementary evidence. None alone establishes RF causation. Registries often lack individual exposure records; symptom reports are subject to selection and attribution bias; clusters require denominators and comparison populations; biomonitoring needs validated, sufficiently specific endpoints.
Develop privacy-protecting linkage between exposure assessment and longitudinal health data, with prespecified hypotheses and independent oversight. Examine genotype-by-exposure interactions where adequately powered and justified. Provide a route for symptom reporting that preserves the distinction between a reported experience and a verified cause. Do not dismiss suffering because attribution is uncertain, or classify every report as proof of exposure-related disease.
Question 11 — Sensitive populations
Yes. Exposure and risk assessment should consider children, pregnancy, older adults, workers, people with existing conditions, and users of implanted medical devices. The reasons differ: developmental timing, exposure geometry, recovery capacity, occupational duration, and device interference are separate questions. A blanket assertion that every child receives more dose per unit body mass under every exposure configuration would be inaccurate.
Genotype and baseline state deserve carefully designed research. The CACNA1C finding motivates replication; it does not identify a proven vulnerable clinical subgroup or justify labeling a variant as defective. [6]
Human prenatal observations deserve targeted follow-up. Bektas and colleagues’ 2018 preliminary study grouped 149 pregnant women by reported mobile-phone use and found higher cord-blood biochemical markers in the group reporting more than one hour daily. Their 2020 placenta and cord-blood study reported oxidative-stress and DNA-damage differences associated with device-use groups. These observational reports should not be treated as randomized exposure tests, established newborn disease, or proof of altered stem-cell programming. Device-use categories are not direct fetal dosimetry; confounding and exposure classification require careful evaluation. Nor should the two reports be counted as independent replication without establishing participant independence. They are distinct from Durdik’s controlled laboratory exposure study. [24–26]
HHS should support prospectively measured prenatal exposure, maternal and delivery covariates, validated cord-blood assays, and longitudinal clinical follow-up. Birth biomarkers do not determine when a change arose during pregnancy or establish an effect on early neural-tube closure.
Use life-stage and tissue-relevant models with adequate dosimetry. The planarian study addresses remodeling in a model organism, not comparative susceptibility in pregnant people or children. Translation requires additional research rather than direct extrapolation.
Question 12 — Evidence below federal RF limits
Assess each study against the actual applicable RF restriction, exposure metric, averaging period, and population. A field strength cannot be called “below limits” without that comparison.
Falcioni and colleagues’ Ramazzini study reported increased heart schwannomas in male rats at its highest 1.8-GHz far-field exposure, with estimated whole-body SAR substantially lower than NTP’s. Its exposure assessment and statistical findings should be evaluated alongside NTP, rather than described as identical experiments or identical findings at every dose. [12]
The CACNA1C study supplies a physiological endpoint under a specific RF exposure; adverse-health significance is a further question. NTP is not below the general-public whole-body limit, and Kim’s 2-mT ELF gene-switch experiment is not evidence of RF harm below federal RF limits. Farrell’s weak ELF experiments likewise concern a different frequency regime. [1,4,6,7]
The proposed 100-µT-peak planarian pilot is not existing evidence. It belongs principally under Question 16. It can examine biological effects in a low-frequency exposure setting expected to produce little bulk heating, but compliance and nonthermal conditions must be assessed, not assumed. Sinusoidal and pulsed exposures require separate characterization. There is no general rule that magnetic fields pass through water with exactly zero attenuation or that the induced electric field is always millivolts per meter; induction depends on geometry and the waveform’s rate of change.
Question 13 — Cumulative exposure patterns
For 5G, measure actual traffic, beamforming, band, source power, and user geometry; antenna count alone does not determine dose. Treat 6G as an emerging set of technologies and avoid assigning one universal waveform or exposure profile. Wi-Fi patterns depend on configuration and traffic, not just beacon repetition.
For satellite communications, distinguish the distant satellite signal from nearby transmitting terminals. For IoT, wearables, smart homes, smart cities, autonomous vehicles, medical devices, and smart meters, record placement, transmission schedule, duty cycle, operating mode, and simultaneous sources. A device’s presence does not establish continuous transmission.
HHS should request comparable deployment measurements across these categories. Exposure accumulation must remain metric-specific: time-integrated power, burst distribution, tissue dose, and overlapping sources are not interchangeable quantities. Our recommendation is better measurement rather than an unsupported assertion that every new network necessarily raises every person’s dose.
Question 14 — Environmental effects
Assess wildlife and ecological endpoints using species-relevant exposure, life stage, behavior, reproduction, and field conditions. Laboratory mammalian cancer evidence should not be presented as an ecological assessment of birds or insects.
Spin-sensitive biological chemistry provides a reason to study magnetic-responsive systems, but does not establish an environmental injury from communications infrastructure. The proposed planarian pilot would test reversion timing in a laboratory organism. It is not evidence of an ecological effect until experiments produce relevant results and their applicability is evaluated. [8,9]
Question 15 — Wireless infrastructure
Consider measurements and validated models at homes, schools, childcare sites, healthcare facilities, and workplaces, including indoors and across relevant operating periods. Evaluate cumulative neighborhood exposure, antenna density, and small cells through their measured contributions, power controls, and geometry rather than simple source counts. Livestock studies should account for feed, handling, housing, infection, and other environmental stressors.
Section 704’s provision, codified at 47 U.S.C. § 332(c)(7)(B)(iv), limits state and local regulation of compliant personal wireless service facilities on the basis of the environmental effects of RF emissions. It does not prohibit federal health research or every discussion of health evidence. RF Safe advocates reconsidering this allocation of authority and strengthening independent federal review. [13]
Question 16 — Highest-priority research gaps and the simple reversion pilot
Our priority is a sustained research program on biological interpretation and recovery: waveform dependence, tissue and genotype susceptibility, and whether repeated exposure can leave a persistent regulatory state that changes later responses. ceLLM proposes these links; S4–Mito–Spin identifies candidate channels, mitochondrial/redox pathways, and spin-sensitive reactions. These are testable proposals, not established explanations of all positive and negative RF studies.
Density gating should be tested by comparing relevant Schwann, glial, skin, and cord-blood cell populations under matched, characterized exposures, measuring the candidate susceptibility variables before outcomes are known. Manipulating a proposed necessary channel or mitochondrial/redox pathway should change the predicted response; measurements of cell identity alone are insufficient. Chronicity requires a separate exposure–washout–challenge design that tests whether a persistent regulatory change predicts altered immune, inflammatory, or metabolic responses after acute effects have resolved.
One low-cost pilot asks a single question: once an altered planarian head is present, does a defined electromagnetic exposure make return to species-typical morphology happen sooner or later? Emmons-Bell and colleagues’ 2015 gap-junction-blockade work supplies the altered-head and subsequent remodeling precedent. It does not establish an RF fidelity assay. [14]
Generate and characterize the altered heads using the established preparation, remove the inducing treatment, and begin exposure only after the phenotype is present. An arbitrary day-ten cutoff is insufficient if formation varies. Randomize comparable starting phenotypes among ambient sham, measured RF/electric-field shielding, continuous-field, and pulsed-field conditions. Candidate added-field settings are 50 or 60 Hz sinusoidal and 217-Hz pulsing at 100 µT peak; exact frequency, pulse width, duty cycle, schedule, and measured field must be fixed before the trial. Those settings are exploratory, not validated biological thresholds.
The only primary outcome is elapsed time to a predefined species-typical head-shape criterion, scored from daily images by an observer blinded to condition. Keep water, temperature, lighting, handling, and enclosure effects comparable. Ordinary Faraday shielding does not reliably remove 50/60-Hz magnetic fields; verify its attenuation and remaining fields. Powered-off coils do not automatically reproduce active-coil heat, vibration, or acoustics.
A reproducible timing difference would show an effect on remodeling under the tested conditions. Faster reversion could reflect destabilization of the induced state, facilitated correction, or another process; slower reversion could reflect stabilization of that state, impeded correction, or another process. Neither direction alone proves higher or lower computational fidelity. Our proposed interpretation is that reversion timing can probe changes in the ongoing physiological runtime. It remains a candidate readout to validate against alternatives.
No detectable difference constrains this pilot’s exposure, endpoint, and observation period; it does not automatically falsify every aspect of ceLLM. Exploratory results should guide a later, prespecified predictive test. Broad signaling research already exists, so we do not claim that all existing work is thermal or genotoxic, or that no non-invasive bioelectric assay exists. We identify this particular exposure-during-reversion timing comparison as a research priority, without claiming an exhaustive novelty search.
Question 17 — Federal coordination
FDA should lead electronic-product-radiation assessment within its authority, publish an evidence and research-gap roadmap, and coordinate exposure-minimization and performance-standard research. The framework originating in Public Law 90-602 includes duties now codified at 21 U.S.C. § 360ii to conduct and support research, evaluate exposure, and develop exposure-reduction techniques. It is a basis for action, not a claim that one particular experiment is legally mandated. [15]
NIH and NIEHS should support independent mechanistic work, replication, developmental endpoints, and longitudinal studies. CDC should develop appropriately validated surveillance and exposure-linked data methods. HHS should coordinate with FCC on standards and infrastructure assessment while retaining independent biological review. Common protocols, accessible dosimetry, declared interests, and publication of null findings would reduce fragmentation.
Question 18 — Additional recommendations
HHS should explicitly distinguish biological interaction, adverse effect, hazard identification, and quantified human risk. Reassess animal evidence and human uncertainty transparently, including IARC’s 2011 Group 2B RF classification and later evidence. Classification is a hazard judgment, not a numerical risk estimate for a particular device. [16]
The D.C. Circuit’s August 13, 2021 decision found deficiencies in the FCC’s explanation for retaining its limits in relation to noncancer evidence and other issues, and remanded for a reasoned response. It did not judicially establish that all compliant exposures are harmful or determine replacement numerical limits. [17]
The Litovitz group’s 1994 chick-embryo study reported that adding a spatially coherent but temporally random magnetic field, at an amplitude comparable to the applied ELF field, reduced the elevated morphological abnormality rate to approximately the control level. This adds a developmental endpoint to the same group’s later biochemical work, rather than independent replication. The experiment measured suppression of abnormalities; it did not measure cancellation of information or removal of field energy. [19]
The Litovitz group’s 1998 ODC study also matters: superimposed temporally incoherent magnetic fields inhibited the measured response to a coherent 60-Hz field. This challenges any universal claim that “more electromagnetic noise always means more biological damage.” It supports studying the exact waveform, interacting fields, and biological task. It does not demonstrate protection from everyday wireless exposure. [3]
These findings motivate a context-dependent hypothesis: an external field may interfere with an imposed biologically active signal or with endogenous signaling, depending on the waveform and biological task. They do not establish ceLLM, a particular transducer, or protection from ambient exposure.
Fund the simple reversion pilot without requiring it to prove the entire theory. Support later pathway and persistence work only as the results justify it. Preserve the urgency of investigation while making every scientific claim precise enough to evaluate.
References
[1] Farrell JM et al. The effect of pulsed and sinusoidal magnetic fields on the morphology of developing chick embryos. Bioelectromagnetics. 1997;18(6):431–438. https://pubmed.ncbi.nlm.nih.gov/9261540/
[2] Effects of low frequency electromagnetic fields on the activity of ornithine decarboxylase in developing chicken embryos. Bioelectrochemistry and Bioenergetics. 1997;43(1):91–96. https://doi.org/10.1016/S0302-4598(96)05174-4
[3] Farrell JM et al. The superposition of a temporally incoherent magnetic field inhibits 60 Hz-induced changes in the ODC activity of developing chick embryos. Bioelectromagnetics. 1998;19(1):53–56. https://pubmed.ncbi.nlm.nih.gov/9453707/
[4] National Toxicology Program. Technical Report 595. https://ntp.niehs.nih.gov/publications/reports/tr/tr595
[5] Mevissen M et al. Effects of radiofrequency electromagnetic field exposure on cancer in laboratory animal studies, a systematic review. Environment International. 2025. https://www.sciencedirect.com/science/article/pii/S0160412025002338
[6] Sousouri G et al. 5G radio-frequency-electromagnetic-field effects on the human sleep electroencephalogram: A randomized controlled study in CACNA1C genotyped volunteers. NeuroImage. 2025;317:121340. https://doi.org/10.1016/j.neuroimage.2025.121340
[7] Kim J et al. Electromagnetic field-inducible in vivo gene switch for remote spatiotemporal control of gene expression. Cell. 2026. Consult linked correction. https://pubmed.ncbi.nlm.nih.gov/41985457/
[8] Ikeya N, Woodward JR. Cellular autofluorescence is magnetic field sensitive. PNAS. 2021. https://pubmed.ncbi.nlm.nih.gov/33397812/
[9] Magnetic resonance control of spin-correlated radical pair dynamics in vivo. Nature. 2026. https://doi.org/10.1038/s41586-026-10282-4
[10] FCC. 47 CFR § 1.1310. https://www.ecfr.gov/current/title-47/section-1.1310
[11] ICNIRP. RF guidelines, 2020. https://www.icnirp.org/en/activities/news/news-article/rf-guidelines-2020-published.html
[12] Falcioni L et al. Report of final results regarding brain and heart tumors in Sprague-Dawley rats exposed from prenatal life until natural death to mobile phone radiofrequency field representative of a 1.8 GHz GSM base station environmental emission. Environmental Research. 2018. https://doi.org/10.1016/j.envres.2018.01.037
[13] 47 U.S.C. § 332(c)(7)(B)(iv). https://uscode.house.gov/view.xhtml?req=granuleid:USC-prelim-title47-section332&num=0&edition=prelim
[14] Emmons-Bell et al. Gap Junctional Blockade Stochastically Induces Different Species-Specific Head Anatomies in Genetically Wild-Type Girardia dorotocephala Flatworms. 2015. https://pubmed.ncbi.nlm.nih.gov/26610482/
[15] 21 U.S.C. § 360ii, Program of control. https://uscode.house.gov/view.xhtml?req=granuleid:USC-prelim-title21-section360ii&num=0&edition=prelim
[16] IARC. Non-ionizing radiation, Part 2: Radiofrequency electromagnetic fields. Monographs Volume 102. https://publications.iarc.who.int/126
[17] Environmental Health Trust v. FCC, No. 20-1025, August 13, 2021. https://media.cadc.uscourts.gov/opinions/docs/2021/08/20-1025-1910111.pdf
[18] HHS. Request for Information on Electromagnetic Fields, Radiofrequency Radiation, and Wireless Radiation Exposure. September 21, 2026; 91 FR 59792–59794. Questions 2–18; deadline October 21, 2026. https://www.govinfo.gov/content/pkg/FR-2026-09-21/html/2026-19252.htm
[19] Litovitz TA, Montrose CJ, Doinov P, Brown KM, Barber M. Superimposing spatially coherent electromagnetic noise inhibits field-induced abnormalities in developing chick embryos. Bioelectromagnetics. 1994;15(2):105–113. DOI: 10.1002/bem.2250150203. https://pubmed.ncbi.nlm.nih.gov/8024603/
[20] RF Safe. Density Gating: Tissue Selective Susceptibility to Modulated Radiofrequency Fields. Position source, not independent mechanistic validation. https://www.rfsafe.com/density-gating-tissue-selective-susceptibility-to-modulated-radiofrequency-fields/
[21] Jyoti J et al. 5G-exposed human skin cells do not respond with altered gene expression and methylation profiles. PNAS Nexus. 2025;4(5):pgaf127. https://doi.org/10.1093/pnasnexus/pgaf127
[22] Impact of in vitro exposure to 5G-modulated 3.5 GHz fields on oxidative stress and DNA repair in skin cells. Scientific Reports. 2025;15:31214. https://doi.org/10.1038/s41598-025-15090-w
[23] SEAWave project. In vitro effects of 5G millimeter wave electromagnetic fields on gene expression in primary human skin cells. Project research summary. https://seawave-project.eu/in-vitro-effects-of-5g-millimeter-wave-electromagnetic-fields-on-gene-expression-in-primary-human-skin-cells/
[24] Durdik M et al. Microwaves from mobile phone induce reactive oxygen species but not DNA damage, preleukemic fusion genes and apoptosis in hematopoietic stem/progenitor cells. Scientific Reports. 2019;9:16182. https://doi.org/10.1038/s41598-019-52389-x
[25] Bektas H, Bektas MS, Dasdag S. Effects of mobile phone exposure on biochemical parameters of cord blood: A preliminary study. Electromagnetic Biology and Medicine. 2018;37(4):184–191. https://pubmed.ncbi.nlm.nih.gov/30156944/
[26] Bektas H, Dasdag S, Bektas MS. Comparison of effects of 2.4 GHz Wi-Fi and mobile phone exposure on human placenta and cord blood. Biotechnology & Biotechnological Equipment. 2020;34(1):154–162. https://doi.org/10.1080/13102818.2020.1725639
[27] Naik S et al. Inflammatory memory sensitizes skin epithelial stem cells to tissue damage. Nature. 2017;550:475–480. Consult linked author correction. https://doi.org/10.1038/nature24271
[28] Wendeln AC et al. Innate immune memory in the brain shapes neurological disease hallmarks. Nature. 2018. https://doi.org/10.1038/s41586-018-0023-4
