Rescission of Regulatory Determinations and Removal of Related Provisions for Four PFAS Substances (PFHxS, PFNA, HFPO-DA (GenX), and the Mixture of These Three PFAS Plus PFBS)

1. The remedy is the open question, and the record names an alternative EPA did not propose. EPA rests the rescission on one reading of section 1412(b)(1)(E): a proposed standard may come no earlier than the final determination to regulate. Four organizations accept that reading for the sake of argument and still oppose the remedy. The National Congress of American Indians (0161) asks EPA to evaluate and publish for comment a sequential cure, a standalone determination followed by a companion rule, that retains the 2024 MCLs; the goDeeper Foundation (0150) says EPA must explain why complete rescission is the only remedy; the Keweenaw Bay Indian Community (0168) asks for the corrected sequence on an expedited schedule; the Massachusetts Department of Environmental Protection (0047) asks EPA to retain the standards or republish them for comment. In NRDC v. Regan (D.C. Cir. 2023) the court held that once EPA makes a final determination to regulate it "shall" regulate and cannot withdraw the determination. EPA's proposal distinguishes that case on the ground that these determinations were themselves unlawfully issued. Whether a court accepts that distinction is the central question on EPA's power to withdraw the determinations; independent anti-backsliding and APA objections bear on the validity and scope of any rescission as well, and the cure the commenters describe is what EPA would be left with if the distinction fails. Confidence: high that the requests are on file and that the precedent governs withdrawal of a determination; moderate on how the court will treat EPA's distinction. Source: organization letters, NRDC v. Regan, No. 20-1335 (D.C. Cir. May 9, 2023).
2. Local governments and state agencies lean opposite ways. Among governments, local governments support the rescission 10 to 3 and state agencies oppose it 8 to 1. The supporting local filers are water systems and utilities, among them the City of Huntsville, Texas (0017), Fairfax Water (0177), Olivenhain Municipal Water District (0160) and the Laurens County Water and Sewer Commission (0086), and their letters argue process, cost and the stepwise statute. The opposing state agencies are environmental regulators, New Jersey (0152), New Mexico (0104), Massachusetts (0047), North Carolina (0179) and the Indiana department on the other side (0091), and the attorneys general coalition (0165) adds the enforcement view. The split is consistent with distinct institutional mandates, systems that pay for treatment against regulators that answer for public health; the groups are small and the letters named are the evidence, not a sample. It signals that a federal rescission would not end the obligations of systems in states that keep or adopt their own PFAS standards. Confidence: high for the counts; moderate for the reading of motives, which rests on the letters named. Source: ComtIQ census, organization letters.
3. The anti-backsliding clause divides the two sides on what counts as a baseline. 27 organizations oppose EPA's finding that section 1412(b)(9) is not triggered, 5 support it. EPA's position is that the baseline is no regulation, because no compliance deadline for the four PFAS has passed, and that an unlawful rule cannot set a protection floor. Chemours (0159) and the National Association of Manufacturers (1021) agree. Iota Intel LLC (0107) answers that a promulgated regulation with a 2029 compliance date is protection in force and that removing it is a revision; the New Mexico Environment Department (0104) calls the zero-baseline reading wrong; the Senate letter (0162) and the attorneys general (0165) say the proposal contravenes the clause. The clause reads "shall maintain, or provide for greater, protection of the health of persons." Which baseline a court adopts decides this issue, and the proposal's own cost memorandum, which counts forgone health benefits, is the evidence opponents cite against the zero baseline. Confidence: high for the positions; low on the outcome. Source: organization letters, 91 FR 29413.
4. Supporters want the rescission and, in several cases, more. 32 comments support the proposal; 27 come from organizations. The American Chemistry Council (0094) asks EPA to rescind the whole 2024 PFAS regulation, not only the four standards. AMWA and AWWA (0061) and the Association of California Water Agencies (0141) support the sequential reading and ask EPA to restart any future determination under it. The Association of State Drinking Water Administrators (0090) asks for time to understand treatment needs if EPA proceeds. A second group supports or accepts the rescission but asks to keep the monitoring and data provisions: the Laurens County Water and Sewer Commission (0086) and Persistence Analytics Group (0519) ask EPA to preserve monitoring and occurrence data, and the Nantucket Tipping Point (0055) asks that data collection not end. Monitoring is the one provision where the two sides come closest. Confidence: high. Source: organization letters.
5. Three petitions carry most of the listed public signatures and count as three comments. Comment 0080, filed by the Natural Resources Defense Council, attaches a petition opposing the rescission with about 12,700 distinct name entries after ComtIQ removed exact duplicates within the list; comment 0170, filed by Environment America, lists 8,174 entries; comment 0083, also Environment America, carries a further list that ComtIQ did not count. ComtIQ counts each petition as one organization comment, consistent with the unit of count throughout this report, and reports the entries here because they are the largest expression of public position in the record. The two counted lists were not checked against each other, so the entries are not a count of distinct people. A listed name is not a verified signatory, and EPA weighs significant comments, not signatures. 18 comments (1.2%) do not address what the rule regulates; they concern sludge incineration, wastewater permits, local contamination disputes and general PFAS concerns, and one of them was read with a side and stays in the totals as read. Confidence: high for the counts; the signatory figures are ComtIQ's count of the lists as filed. Source: ComtIQ census, analyst reading of 0080, 0170 and 0083.

Position by unit of count. A comment is one count; a campaign is one text sent many times, and the record does not show whether by one person or many; one organization may file twice. Each line counts the same record a different way. The deduplicated units are not a count of distinct people or entities: text similarity does not establish common authorship.

Gap recorded by the census (attachments listed without file): 2 comments, both from individuals, reported here in aggregate only.

Individuals are reported only in aggregate. The record holds 124 organization letters. All 28 that are longer than the coding system's reading limit were read in full by ComtIQ's reading assistant, a second model reading the whole letter; the analyst checked each reading against the letter and confirmed or corrected it. Organizations are named in this report where a finding or an argument rests on them; the list of all filers with their stance is in the data file organizations.csv.

Coalitions. The largest institutional coalition is the letter of 19 attorneys general, from 18 states and the District of Columbia, led by North Carolina (0165): North Carolina, California, Connecticut, New Jersey, Arizona, Colorado, Delaware, the District of Columbia, Hawaii, Illinois, Maine, Maryland, Massachusetts, Minnesota, New York, Oregon, Rhode Island, Washington and Wisconsin. Earthjustice, the Natural Resources Defense Council and the Southern Environmental Law Center filed jointly (0842), with ten further comments carrying only the references cited in that letter, and a broader Earthjustice-led coalition filed separately (0110). Thirty-four craft breweries in twenty states signed one letter (0180), led by Sierra Nevada Brewing Co., Lakefront Brewery, Half Acre Beer Company and Port City Brewing Company. AMWA and AWWA, the two national associations of water utilities, filed one joint letter in support (0061). Two tribal governments, the Keweenaw Bay Indian Community (0168) and the Snoqualmie Indian Tribe (0113), and the National Congress of American Indians (0161) filed, all opposed. The letter recorded as "United States Senate" (0162) carries one signature in the docket text, Senator Adam Schiff of California; press reports describe a letter from 26 senators to the Administrator, so the docket understates the congressional position. ComtIQ did not find a letter from the House in the record.

Themes, by number of comments carrying them (a comment can carry several).

Argument types, by number of comments.

Provisions, coded for organizations and law firms only. Unit: comments, so an organization that filed twice counts twice here; the section on disputes below counts each organization once. Oppose, support and modify here are a stance on the provision, not the comment's position on the proposal: a commenter can support the rule and ask to remove one safe harbor.

The strongest arguments, by side.

For the rescission. The statutory text, as EPA now reads it: section 1412(b)(1)(E) says EPA shall propose a standard "not later than 24 months after the determination to regulate," and the determination the statute refers to is the final one, published after comment on a preliminary determination. AMWA and AWWA (0061) and the Association of California Water Agencies (0141) add the practical point that collapsing the two steps denied the public a chance to comment on whether to regulate before commenting on how. Chemours (0159) supplies the anti-backsliding answer: rescinding a rule issued without authority is not a revision of a lawful standard, and the 2024 standards had no compliance date in effect. The National Association of Manufacturers (1021) argues the same and adds occurrence and hazard-index objections as confirmation. The weakness on this side is remedy: none of these letters explains why the defect, if it exists, requires withdrawing the determinations rather than re-proposing the standards.

Against the rescission. The text, read the other way: "not later than" sets a deadline, not a starting gate, and nothing in the subsection forbids an earlier proposal; the New Jersey Department of Environmental Protection (0152) says the provision expressly authorizes concurrent publication, and the Cape Fear Public Utility Authority (0137) says EPA has proposed standards with final determinations before. The New Mexico Environment Department (0104) argues harmless error under 5 U.S.C. 706: the public did comment on both the determination and the standards. The American Association for Justice (0088) answers EPA's reliance on Loper Bright: the end of deference does not oblige an agency to reverse its own prior actions. On remedy, NRDC v. Regan and the "shall" in section 1412(b)(1)(E) support the NCAI (0161) and Keweenaw Bay (0168) position that the determinations stand and the cure is procedural. On anti-backsliding, Iota Intel LLC (0107) makes the sharpest point: a promulgated standard with a 2029 compliance date is protection, and the proposal's own memorandum counts the health benefits that rescission forgoes. On scope, the Earthjustice coalition (0110), EDGI (0118) and the Kentucky Resources Council (0183) argue that an agency reversing a rule must consider reliance interests and the consequences of reversal, and that a notice limiting comment to legal grounds does not relieve EPA of that duty. The weakness on this side is that the D.C. Circuit has not yet said whether EPA's reading of the sequence is wrong; two denials of summary vacatur say only that the question is not simple.

What a court would weigh. The sequence question is a pure question of statutory text that the court will decide without deference (91 FR 29413; NRDC v. Regan, No. 20-1335, D.C. Cir. May 9, 2023; petitions for review under SDWA section 1448, 42 U.S.C. 300j-7). The remedy question turns on whether the 2024 determinations, which followed a preliminary determination and public comment, were themselves defective or only the standards proposed alongside them were. If the determinations stand, NRDC v. Regan binds EPA to regulate, and the cure is to re-propose. If they fall, the anti-backsliding and reliance arguments remain, and EPA's zero-baseline reading is the next contested step.

Provisions ranked by the number of organizations asking to change or remove them. Unit: organizations, each counted once per provision. Each row is one organization, what it asks as the coding system summarised it, and the comment to open; the summary is a reading, not a quotation. Read the letter before quoting it.

rescind-reg-determinations: Rescission of the final regulatory determinations to regulate PFHxS, PFNA, HFPO-DA and the mixture of these with PFBS (Index PFAS) as unlawfully issued.

Organizations opposing it 71, asking to modify it 7, supporting it as proposed 26.

96 more organizations on this provision are in organizations.csv and the reading copy.

remove-mcls-mclgs-141-61: Removal of the MCLGs and MCLs for PFHxS, PFNA, HFPO-DA and the hazard index from 40 CFR 141.61(c) and the associated best available technology entries in 141.61(d) and 142.62.

Organizations opposing it 62, asking to modify it 1, supporting it as proposed 12.

67 more organizations on this provision are in organizations.csv and the reading copy.

legal-reading-1412-b-1-e: EPA's new interpretation of SDWA section 1412(b)(1)(E) that 'determination to regulate' means only the final determination, so a proposed NPDWR cannot be issued concurrently with a preliminary determination.

Organizations opposing it 26, asking to modify it 3, supporting it as proposed 22.

43 more organizations on this provision are in organizations.csv and the reading copy.

no-anti-backsliding: EPA's finding that rescission does not trigger SDWA section 1412(b)(9) anti-backsliding because the rule was unlawful, and alternatively that it maintains protection since no compliance deadlines have passed.

Organizations opposing it 27, asking to modify it 1, supporting it as proposed 5.

25 more organizations on this provision are in organizations.csv and the reading copy.

remove-monitoring-compliance-provisions: Removal or revision of monitoring, trigger-level, compliance and violation provisions for these PFAS in 40 CFR 141.902, 141.903 and 141.905, and of related analytical method and performance evaluation entries in 141.901.

Organizations opposing it 18, asking to modify it 3, supporting it as proposed 6.

19 more organizations on this provision are in organizations.csv and the reading copy.

legal-grounds-only-scope: EPA's choice to propose the rescission solely on legal grounds and treat comments on health risks, costs or occurrence findings as outside the scope of the rulemaking.

Organizations opposing it 13, asking to modify it 1, supporting it as proposed 0.

6 more organizations on this provision are in organizations.csv and the reading copy.

By provision.

rescind-reg-determinations. The question is whether a final determination to regulate, made in April 2024 after comment on a preliminary determination, can be rescinded because the standards were proposed alongside the preliminary one. 71 organizations oppose, 26 support, 7 ask for a change. EPA is likely to rescind as proposed; it has argued in court that the determinations are unlawful and it has not offered a middle course. The middle course is in the record, from the NCAI (0161), goDeeper (0150), Keweenaw Bay (0168) and MassDEP (0047), and EPA's response-to-comments document will have to meet it.

remove-mcls-mclgs-141-61. The standards fall with the determinations; 62 organizations oppose, 12 support. The International Carwash Association (0173) asks a narrower thing, a preamble statement that removal undercuts categorical industrial PFAS sampling, which shows how the standards reach beyond drinking water into pretreatment. The Institute of Hazardous Materials Management (0030) asks EPA to stay rather than remove. Likely rescinded with the determinations.

legal-reading-1412-b-1-e. The core legal dispute, 26 against, 22 for. EPA will keep its reading in the final rule; the court decides whether it is the best one. The letters for the reading come from the regulated water sector, the letters against from state regulators and attorneys general.

no-anti-backsliding. 27 against, 5 for. EPA will keep both findings, that the clause does not reach an unlawful rule and that the baseline is no regulation. The second is the weaker one, because it rests on the absence of a compliance deadline rather than on the text of the clause, and the proposal's own cost memorandum counts forgone benefits. Expect this to be a lead argument in any petition for review.

remove-monitoring-compliance-provisions. 18 against, 6 for, 3 ask for a change, and the asks cross sides: the Laurens County commission (0086) and Persistence Analytics (0519), which accept the rescission, join Harpeth Conservancy (0172), Culligan (0696), ISEE (0100) and New Mexico (0104) in asking to keep monitoring or data collection. The initial monitoring deadline of April 26, 2027 is already funded by many systems. This is the provision where a final rule could move without conceding the legal question; EPA's statement that it will re-evaluate the four PFAS gives it a reason to keep occurrence data flowing. ComtIQ still judges a monitoring carve-out unlikely, because the proposal removes 141.902 to 141.905 as a package and UCMR 5 already supplies national occurrence data.

legal-grounds-only-scope. 13 against, none for. EPA will keep the framing. The exposure is in the response to comments: if the final rule dismisses health, cost and reliance comments as out of scope while its own memorandum counts forgone benefits, the APA argument that the agency failed to consider an important aspect of the problem is set up for the court.

Letters with near-identical text are grouped by a similarity detector whose thresholds are fixed before any docket is read. Members take the reading of the group's representative text.

Campaign 1. Its reading rests on: "As a stakeholder, I strongly oppose your proposal to rescind regulatory determinations and related provisions for four PFAS substances (PFHxS, PFNA, HFPO-DA (GenX), and the mixture of these three PFAS plus PFBS)." Themes: anti-backsliding-principle, public-health-pfas-exposure, vulnerable-populations.

Campaign 2. Themes: polluter-pays-liability, public-health-pfas-exposure, science-and-evidence-basis.

Campaign 3. Themes: public-health-pfas-exposure.

Campaign 4. Themes: public-health-pfas-exposure.

Campaign 5. Themes: public-health-pfas-exposure, science-and-evidence-basis.

Campaign 6. Themes: anti-backsliding-principle, public-health-pfas-exposure.

Campaign 7. Themes: deregulation-policy-priorities, public-health-pfas-exposure.

Campaign 8. Themes: public-health-pfas-exposure, science-and-evidence-basis.

Campaign 9. Themes: mixtures-cumulative-exposure, public-health-pfas-exposure, vulnerable-populations.

Campaign 10. Representative text: EPA-HQ-OW-2025-0654-0115. Its reading rests on: "The Agency should not proceed with the Proposed Rule to Rescind Regulatory Determinations and Regulations for PFHxS, PFNA, HFPO-DA (GenX), and the mixture of these three PFAS plus PFBS (EPA-HQ-OW-2025-0654)." Themes: anti-backsliding-principle, private-wells-and-unregulated-sources, public-health-pfas-exposure.

Campaign 11. Themes: public-health-pfas-exposure, regulatory-stability-and-trust, vulnerable-populations.

Campaign 12. Themes: public-health-pfas-exposure.

Campaign 13. Themes: anti-backsliding-principle, deregulation-policy-priorities, public-health-pfas-exposure.

Campaign 14. Themes: anti-backsliding-principle, public-health-pfas-exposure, statutory-authority-and-courts.

Campaign 15. Themes: public-health-pfas-exposure.

The largest campaign (84 letters, all anonymous) asks EPA not to rescind the determinations and the standards, in one template sentence, and cites public health, the anti-backsliding principle and vulnerable populations; the record names no sponsor. The only campaign with a named sponsor is the NC Sierra Club's (5 letters). The three petitions (0080, 0170, 0083) are not campaigns under ComtIQ's detector, which groups repeated texts across comments; each is one comment carrying a list of signatories, and the lists are reported in finding 5.

Not applicable at the proposed-rule stage. At the final rule, each significant issue above is matched to the preamble's response and unanswered issues are flagged.

What was said about the proposal outside the docket: law firm analyses, specialist press, statements and coverage that ComtIQ's analyst read and kept. These sources are not comments; nothing here is counted with the record above. A source marked 'filer' was published by an organization that also filed in the docket: it restates a party's position and is not independent corroboration. The summaries are ComtIQ's; read each source before relying on it.

Outside discussion matches the record on the legal frame and adds what the record lacks: timing. The law firm alerts (Morgan Lewis, Thompson Hine, Ballard Spahr, Pillsbury) read the proposal as a procedural correction designed to avoid the anti-backsliding clause and expect litigation from both sides; the Harvard Environmental and Energy Law Program adds that EPA's new reading would slow every future SDWA rulemaking by adding a second comment round. On the supporting side, AMWA's own mid-year report confirms the position in its joint letter. The comment docket does not show what was argued at the July 7 hearing, whose recording EPA posts on the rule's page. Nor does it show the meetings held at the Office of Information and Regulatory Affairs while the proposal was under review; reginfo.gov lists three for RIN 2040-AG53, on March 27, April 9 and April 20, 2026, requested according to those records by the Environmental Working Group, by the American Chemistry Council with Chemours in attendance, and by the Alliance of Nurses for Healthy Environments. The records show that the meetings took place and who attended; they do not show what was decided. Nor does the record show the D.C. Circuit's view, which will arrive in American Water Works Association v. EPA, No. 24-1188, argued September 18, 2026.

Each judgment carries a likelihood term and band (ICD 203), a confidence and the test that settles it, and goes into the forecast log. The bands are the analyst's estimates, not frequencies from the census or outputs of a validated model; they rest on the statute, the proposal, the court record and what EPA has said. The number of commenters on a side is context, not the basis.

What would change this view: a D.C. Circuit opinion in AWWA v. EPA that reaches the sequence question would move S2, S4 and S7 sharply in either direction; a supplemental notice from EPA on monitoring would raise S3; a final rule that answers the NCAI and goDeeper letters with a severability or re-proposal plan would raise S4.

What the record means for each reader, the provision and forecast behind it, and when to act.

Dates: 2024 rule published April 26, 2024 (89 FR 32532), with initial monitoring due April 26, 2027 and MCL compliance due April 26, 2029; a separate 2026 proposal would let systems seek until 2031 for PFOA and PFOS, and it is a proposal; proposal published May 20, 2026 (91 FR 29413); hearing July 7, 2026; comments closed July 20, 2026; AWWA v. EPA argued September 18, 2026, decision pending; no final rule found as of November 7, 2026. A petition for review of a final rule must be filed within 45 days of its promulgation (SDWA section 1448, 42 U.S.C. 300j-7), and filing does not stay the rule. A final rule takes effect on the date its text sets, whatever the state of the litigation, unless a court orders otherwise. Watch the Federal Register and the Unified Agenda for RIN 2040-AG53 and the D.C. Circuit docket for No. 24-1188.

Collection. Every posted comment in docket EPA-HQ-OW-2025-0654 was pulled through the Regulations.gov v4 API and stored with a dated snapshot, with every attachment the API listed except 2 that the API lists but does not serve (comments 1062, 1260), read from the comment field alone; attachments were read as PDF, Word or plain text, scanned PDFs through OCR, and image-only files recorded as comments without words. Census generated Nov. 8, 2026.

Processing. Exact and near duplicates grouped and coordinated campaigns detected as described above. One representative per campaign was coded; 1,377 comments were read one by one, holding 2,345 distinct texts once each attachment is counted on its own, and campaign members inherit their representative's reading.

Analysis. Each unique text was read and coded by ComtIQ's AI coding system against a fixed codebook, returning position, commenter type, themes, argument types, provisions for organizations, and the sentence each judgment rests on. 39 comments were flagged as uncertain by the coding system (unclear stance, attachment-only text or a reading it doubted); 35 are longer than the reading limit (see the limits below), and 33 are both. All 39 flagged readings were then read in full and confirmed or corrected, so none is open. The flags are reported as coverage and the reading stands unless the analyst corrected it. Campaign members inherit the reading of their representative.

Measurement. The unit is the comment read directly: 1,377 of them, campaign copies once (the 2,345 distinct texts above count comment fields and attachments separately). A simple random sample, not stratified, of 15% (207; the share is set by docket size, all of a docket up to 100 comments, half up to 500, 15% up to 2,000, 5% above) was read a second time, blind, at a stronger reading level. Agreement between the two readings, not accuracy against a human: position in three values (support, oppose, other) 99% (95% interval 96% to 99%, n=207); organizations 100% (95% interval 87% to 100%, n=24); individuals, which here include anonymous and unclear commenters, 98% (95% interval 96% to 99%, n=183); commenter group 97% (95% interval 94% to 99%, n=207). The group sizes come from the draw, not from a quota. Intervals are Wilson intervals with the finite population correction; the population is the 1,377 comments read directly, and for each group its share of them as estimated from the sample. These figures cover position and commenter type only; themes, argument types and provisions are not measured.

0 posted comments carry no readable text. 35 comments are longer than 40,000 characters; the coding system read each up to that limit, and each was then read in full, 1 by the analyst and 34 by ComtIQ's reading assistant, whose readings the analyst checked; each reading was confirmed or corrected. Themes, argument types and provisions are automated readings not yet measured against a second reading; they are reported as counts, not as findings, unless an analyst confirmed them. No comment is labeled as written by AI; nothing here speaks to authorship.

The data behind the tables (organizations with their stance, campaigns, position by commenter type) is published with this report as CSV files.