Generated 2026-10-21 from the project record. This page lists the correction directives open against the live properties, or not yet tied to one, by record id and date filed. The live properties are the ALETHEIA sites, holisticquality.io and the Dimensional Literacy hosts. Findings not yet placed in a directive are not listed.
- D-631
- State: partial · date filed: 2026-08-21
- In progress: 509 out-of-date product hazard levels have been recalculated. For 99 more, a recalculation gives a lower level than the one shown, or none at all for 2 of them, and 79 of the 99 are in children's contexts; they keep their current level until a review of how product levels are worked out decides.
- D-868
- State: partial · date filed: 2026-09-23
- In progress: our product calculation reads no ingredient amounts and divides the most hazardous ingredient's contribution by the number of ingredients, which can understate a product's hazard. The planned default counts the worst ingredient at full strength unless a sourced concentration supports less; the options are measured first and no stored level changes before review.
- D-916
- State: partial · date filed: 2026-09-29
- Planned: fix records whose identity is wrong or mixed: names that do not match the CAS number, synonyms that belong to other substances, data left over from a different substance, and structure strings with no recorded source.
- D-928
- State: partial · date filed: 2026-09-29
- Planned: add the EU's workplace exposure limits from its 2009 and 2017 directives, marked as binding or indicative, together with the EU asbestos limit and the prohibitions under the chemical-agents directive.
- D-929
- State: partial · date filed: 2026-09-29
- Planned: fix workplace-limit records where template text was copied across related substances, where "Not applicable" is used as if it were a finding, and several single-substance errors, including a stricter limit restored to the body that set it.
- D-934
- State: queued · date filed: 2026-09-29
- Planned: every number that decides a hazard level we publish must come from an authority, follow from a stated method, or be disclosed and justified as our own convention. Any number that meets none of these will be replaced or removed.
- D-935
- State: queued · date filed: 2026-09-29
- Planned: many reference values on compound pages credit the EPA's IRIS programme for figures IRIS never set or states differently, often repeated in a template sentence. Each will be corrected to the setting body's value, re-credited, or have its credit withdrawn with a note; five IRIS values looser than the EPA pesticide office's are made stricter first.
- D-936
- State: queued · date filed: 2026-09-29
- Planned: other daily-intake and reference-dose values that we credit to the EPA will be credited to the body that actually set them, or replaced with a sourced value or a marker.
- D-937
- State: queued · date filed: 2026-09-29
- Planned: our hazard calculation ignores some cancer classifications because it has no mapping for the agency, and reads one qualified "not likely" cancer finding as a plain negative, which can understate the hazard. Each agency will be mapped or declared not a cancer source, with a check for new ones, and the affected records' stored levels recalculated from their own data where a rule exists, and otherwise labelled as authored with no recorded basis.
- D-938
- State: queued · date filed: 2026-09-29
- Planned: many daily-intake and reference-dose entries on our compound pages hold placeholder text that names no EPA body. Each will become a sourced value or a note saying why no value is given; an unsourced figure credited to an agency loses that credit or is replaced, and label maximums and upper limits are added where a source gives them.
- D-942
- State: queued · date filed: 2026-09-29
- Planned: a product's hazard level is now a fixed 60/40 blend of its worst and average ingredient, each diluted by the number of ingredients. It will instead compare each ingredient's worst-case dose with a reference value, showing an ingredient with no reference as unassessed, never safe; the work stops for review if most ingredients still lack reference values.
- D-946
- State: queued · date filed: 2026-09-29
- Planned: no compound page shows a line of legally binding limits. A new "Legal limits" line will show each limit's source, how it applies and whom it binds ("not stated" where the source does not say), with recommendations labelled apart; our US drinking-water limits will be checked against the federal regulation, missing ones first.
- D-948
- State: queued · date filed: 2026-09-29
- Planned: we will build a table of which rules we cover, by country and instrument, including the EU REACH Candidate List. Missing rules will then be added in this order: EU cosmetics restrictions, US children's-product rules, then the children's-product lists of Washington, Oregon and Vermont.
- D-955
- State: partial · date filed: 2026-09-29
- Planned: almost no compound page links a checked source document, and 323 pages show no Sources section. We will find, confirm and link the documents our sources name, label database links as "where to look", mark unchecked values meanwhile, and within this work's scope correct values to their source or withdraw those no source can be found for.
- D-968
- State: partial · date filed: 2026-09-30
- Planned: 824 of 959 materials and 885 of 1,287 products link no source document, and the 1,055 regulatory entries on materials cite nothing and repeat a few template sentences. After checking samples first, we will link the documents behind these statements and withdraw what no source supports, which may mean withdrawing most template entries.
- D-971
- State: partial · date filed: 2026-09-30
- In progress (measured 2026-10-12 in our data; it reaches the site at the next update): 543 of the 548 sets of hazard codes (the GHS classification) on our compound pages that are the EU's official harmonised classification now cite the EU's own list (CLP Annex VI) instead of PubChem, a database. The other 5 harmonised sets keep their codes and their PubChem citation, with a note, because their entries in the EU list could not be matched one to one. 712 sets are classifications that companies notified to the European Chemicals Agency; they still name PubChem and are counted as not yet sourced. 5 sets have no classification type. How the notified sets will be sourced is still being decided.
- D-972
- State: in-progress · date filed: 2026-09-30
- Planned: many cancer classifications on our compound pages name the classifying body without linking its document, and many regulatory-list entries carry no source. Classifications, listings and "not on this list" statements in this work's scope will be checked against the body's own document or list, and linked, corrected or withdrawn with a note.
- D-973
- State: queued · date filed: 2026-09-30
- Planned: most established exposure limits on our compound pages (workplace, air-quality and similar) name the body that set them without linking its document. Once other work has corrected the values, each limit in scope will be compared word for word with the body's own document and linked, corrected (stricter values first) or withdrawn if no document states it.
- D-974
- State: in-progress · date filed: 2026-09-30
- Planned: most hazard blocks on our compound pages (such as skin sensitisation) do not cite the body that set their values, and some values appear to be our own inferences. 58 "not a skin sensitiser" statements contradict the compound's own hazard codes, understating the hazard, and are corrected first; the others will be sourced, marked "not checked", labelled as model estimates, or withdrawn.
- D-975
- State: partial · date filed: 2026-09-30
- Planned: most no-effect and lowest-effect doses on our compound pages link to a database entry, not the study behind them; 1,446 lethal-dose (LD50) items cite a source that resolves to nothing, some lethal-dose values came from a hazard category, and 179 no-effect doses were estimated from a lethal dose. After sampling, each will be linked, corrected or withdrawn.
- D-976
- State: partial · date filed: 2026-09-30
- In progress (counted 2026-10-12 in our records; not yet on the site): of 15,167 identifier values in our compound records, 4,266 structure codes and names now cite PubChem's computed record. A further 6,755 registry identifiers (PubChem, CompTox, ChEBI and UNII identifiers and similar) are assigned by a named registry, but for most of them we have not yet attached that citation to the records, and agreement with the registry was sampled, not checked value by value. 4,146 identifier values have no registry citation, most of them cross-references or names that no registry issues. Among property values, 1,417 computer-model predictions are recorded as model estimates published by EPA CompTox, 10,362 values with no source have been replaced by a not-checked marker, and 8,609 still name no source. Each remaining property value will cite a regulator document, a named handbook or a published study, or be withdrawn.
- D-983
- State: partial · date filed: 2026-10-01
- Planned: where a regulator's own list or document has not yet been archived (for example the IFRA Standards, the US drug-schedule list, and several EU and FDA lists), archive it from the regulator's own site, then point the classification rows that rely on it at that document. Rows for which no such document exists will be listed with the reason, never pointed at a guessed address.
- D-991
- State: queued · date filed: 2026-10-08
- Planned (counted 2026-10-13 in our records): one compound record is named iron oxide, but its formula (Cr2O3) and almost all the other names in its own list are those of chromium(III) oxide. Ten material and product records point to it; in four of them a line calls it iron oxide (for example the colorant in calamine lotion and an ingredient of a laser printer toner cartridge). Those products may be described with the toxicology of the wrong oxide, and correcting that could move a product's band either way. The record's name will be re-read from an identity registry for its CAS number, then the record renamed and each ingredient line pointed at the correct iron oxide record where its own text names iron oxide. The 89 compound records created from material lines will also be checked, name against CAS number.
- D-993
- State: queued · date filed: 2026-10-08
- Planned (counted 2026-10-13 in our records): 1,393 compound records serve an OSHA permissible exposure limit of "Not established" with the basis given only as "N/A", so nothing says which OSHA tables or substance-specific standards were searched. They are part of a larger group of 1,415 records whose OSHA limit reads or begins with "Not established", also listed on this page under the check of OSHA's dust limit. Examples are glyphosate, bisphenol A, caffeine and sodium hypochlorite (bleach). For many of these "Not established" may be correct; the gap is that the record does not say what was searched, and where a limit does exist the record is less protective than it should be. Each substance will be checked against OSHA's own tables (29 CFR 1910.1000) and its substance-specific standards, with a known case run as a control. Where the absence is confirmed, the record will say what was searched; where a limit is found, it will be added; where neither can be shown, the record will say the limit was not checked.
- D-998
- State: queued · date filed: 2026-10-08
- Planned (counted 2026-10-13 in our records): none of our 1,287 product records has a field for which version of the product it describes (barcode, formulation date or lot). A maker can change a formula, and a record could then go on describing an ingredient list that is no longer sold, with nothing to mark it out of date. We will add a version note to every product record, set to "version unknown" unless a public, archived source supplies the barcode or formulation date. Where such a source dated later shows a different ingredient list, the old record will be marked as superseded and kept readable and labelled, so the old formula is no longer presented as current. Evidence on a record of unknown version will be able to raise or hold a value, never lower it.
- D-1005
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): on 22 compound records, a tolerable intake or reference dose may differ from what its issuing body publishes, and in most cases may be less protective. Two cases show it from our own records: 5 PFAS records serve 1.86 ng/kg body weight per day credited to EFSA 2020 while the basis they quote, a weekly intake of 4.4 ng/kg for four PFAS together, works out to about 0.63 ng/kg per day; 8 more PFAS records serve the same 1.86 credited to an EFSA 2020 group assessment that may not cover them. Three mercury records for methylmercury compounds serve EFSA's inorganic-mercury intake of 0.57 µg/kg per day, while their own notes give 0.19 µg/kg per day for methylmercury; a fourth, elemental mercury, serves the same inorganic figure and is also being checked. The others concern PFAS reference doses on five of those PFAS records, inorganic arsenic, T-2 toxin, and two intakes (ochratoxin A and chlorpyrifos) that their issuing body may no longer publish. Each value will be re-read from the issuing body's own document, then corrected to the figure it prints, withdrawn where none is printed, or left as it is where it already matches.
- D-1009
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): the hazard classifications behind our risk bands may be incomplete or mislabelled in places, and some may be less protective. Two cases from our records: 43 compounds listed under California Prop 65 serve fewer listing types than our archived copy of the state's list carries for the same CAS number (ethylene oxide serves "cancer"; the archived list also carries "female", "developmental" and "male"); and 104 of the 115 compounds with an IARC Group 1 entry serve no concern level in their carcinogenicity entry. Each value will be re-read from its issuing body's own document, then corrected to the figure it prints, withdrawn where none is printed, or left as it is where it already matches.
- D-1010
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): we have 18 open findings about regulatory statements we serve (limits, bans and listings) where the issuing body's own text is not yet on file or has not yet been read against the statement; each statement may differ from what that body prints, in either direction. Two cases from our records: the phrase "SVHC/banned/restricted in EU" is served on 14 entries of our fragrance catalogue, and we have not yet checked those entries against the Regulation; and one material record states an EU limit of 0.01 mg/kg for a food-contact substance, while none of the record's seven sources mentions EU Regulation 10/2011. Each statement will be re-read from the Regulation or notice itself, then corrected to what it prints, withdrawn where no text is on file, or left as it is where it already matches.
- D-1011
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): some hazard wording on compound records is boilerplate or may have no document behind it. In our records 1,589 placeholder lines say "not yet compiled" across 1,250 records; five records carry the short regulatory note "safe chemistry" (for example "No SVHC; no restrictions; safe chemistry"); and the sentence "Low_to_moderate risk based on available hazard data." appears on 43 records, 35 of which have an empty hazard-code block (we will check whether the sentence rests on other data, such as listings). Each sentence will be read against its source, then withdrawn to "not checked" where no source states it, sourced where one does, or left as it is where it already matches.
- D-1012
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): some identifiers and structure descriptors on compound records may belong to another substance, which would make the descriptors served for those records wrong. In our records 22 PubChem numbers are each held by more than one record that is not an alias (31 if aliases are counted); one number is held by four records for different forms of hexachlorocyclohexane (lindane, alpha-HCH, beta-HCH and a mixed-isomer record), and the tolyltriazole record lists among its synonyms the name of an acid, not a triazole. Separately, 565 structure strings are written differently in two places they are stored (in a sample of eight we read, all were the same structure in a different notation), and 13 exact masses differ between two places they are stored. Each identifier will be re-read from its issuing registry, then corrected, moved to "not checked" where the registry on file cannot answer, or left as it is where it already matches.
- D-1014
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): some dose, reference and occupational values we serve may differ from what their issuing body publishes, and some child and infant bands may sit below the adult band, which would be less protective. Toluene, 2,4-dinitrotoluene and the food additive BHT all serve the same acceptable daily intake, 8.6 mg/kg body weight per day credited to JECFA 1999; the same block is on 5 of the 26 records that carry one. On our ordering of risk-band names, stored child bands are below the adult band in 83 of 910 comparable pairs, and infant bands in 43 of 1,172. Each value will be re-read from the issuing body's own document, then corrected to the figure it prints, withdrawn where none is printed, or left as it is where it already matches.
- D-1015
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): some material and product statements may not have a document behind them. 260 of our 959 material records serve a decomposition temperature, for example a single figure for a broad category such as graphene and carbon nanomaterials. 133 products have a derived band of "insufficient data" while still carrying an authored risk level, and 146 product regulation entries are written as plain text instead of a structured entry with a source field. 27 of 1,724 testable composition lines link a compound whose CAS number is not among those printed on the line (for example a paraformaldehyde line in a flow cytometry reagents material is linked to the formaldehyde record); each will be read to decide whether the link is a deliberate stand-in or a mistake. Each statement will be read to its source, then corrected, withdrawn, or left as it is where it already matches.
- D-1021
- State: queued · date filed: 2026-10-13
- Planned (counted 2026-10-12 in our records): 1,415 compound records serve an OSHA limit that reads or begins with "Not established" (1,393 of them with the basis "N/A", also listed on this page under the check of OSHA absence statements). Some of them may be inert or nuisance dusts that OSHA's catch-all dust limit (29 CFR 1910.1000(f)) covers, so the absence statement may be incomplete and less protective. Separately, five records (pendimethalin, tributyltin oxide and three microcystins) may not carry the strictest intake value published by any authority, and six organotin records each serve a tolerable daily intake whose own basis line describes an inorganic tin figure. Each value will be re-read from the issuing body's own document, then corrected to the figure it prints, withdrawn where none is printed, or left as it is where it already matches.
- D-1023
- State: queued · date filed: 2026-10-14
- Planned (checked 2026-10-14 in our records): a review found four sentences on four product records that the source cited for them does not state, for example one about dogs sleeping on flame-retardant-treated pet beds and one saying an ingredient was banned together with oxybenzone, and a source note on the pet-bed record that describes a different paper. Each sentence, with any copy of it on the same record, will be withdrawn and the note removed; finding a source that states them is later work. Separately, two infant products give the same attributed figure as 180 ppb arsenic in some places and 177 ppb in others, and one of our research pages prints 180 ppb; nothing changes there until we have archived the 2021 US House of Representatives staff report on baby food and read which figure it gives, for which product and for which metal.
- D-1025
- State: queued · date filed: 2026-10-14
- Planned (counted 2026-10-14 in our records): earlier work withdrew a set of template hazard sentences from the per-context safety summaries on compound records, but copies survive elsewhere on the same records: 11 on 8 records (in an older second copy of those summaries, in two intake notes and in one note on a safer alternative), one in a safety-advice note, and 85 in change-history notes that our public pages do not show, most of which will be kept as history. On the insect repellent DEET, six statements give a 30% concentration limit for children or a 10-30% range: our archived FDA text gives the 30% limit as the American Academy of Pediatrics' recommendation, our archived EPA page states no limit, and a statement crediting a range and safety to an EPA assessment, which the EPA page does not support, will be withdrawn. The other copies will be made to match the summary they repeat, re-based on their true source, or withdrawn, with the earlier text kept on the record.
- D-1026
- State: queued · date filed: 2026-10-14
- Planned (counted 2026-10-14 in our records): on 40 compound pages as built in our records, the line that tells a reader where the hazard classifications are published is labelled as where to look, not a source and not a regulatory ruling, but the sentence that introduces it still calls it the codes' source. That sentence will be changed to say that the line shows where to look for those codes. No hazard code, signal word or safety band changes.
- D-1031
- State: queued · date filed: 2026-10-15
- Planned (counted 2026-10-15 in our records): three compound records that are alternative-name entries for other compounds still carry 24 regulatory-listing values (whether the substance is on a list such as Canada's Domestic Substances List, or was detected in a US national health survey) and three blocks of computer-model estimates, none with a source. The same kinds of value were already replaced by a "not checked" label on other records, but these three records were skipped; they appear in the safety site's published data files, not on a compound page. Each value will be set to "not checked"; for the listings, the main records these entries point to already show "not checked" or no such value.
- D-1036
- State: queued · date filed: 2026-10-15
- Planned (counted 2026-10-15 in our records): 2,280 study dose values on 492 compound records (no-adverse-effect, lowest-adverse-effect and median lethal doses) name the European Chemicals Agency as their source, 2,007 of them its study database, and we cannot open that source to check them. We will decide whether they stay, labelled as a pointer to where to look, are withdrawn, are first checked against a sample of pages saved from the agency's own site, or whether the agency is asked for permission; how a withdrawn dose value is shown must also be decided, because no form for it exists yet. Nothing changes until then, and any withdrawal keeps the value's place and records the earlier value.
- D-1037
- State: queued · date filed: 2026-10-15
- Planned (counted 2026-10-15 in our records): 15 compound records keep an older second copy of their per-context safety summaries (adult, child, dog, cat and aquatic life) beside the maintained one, and our published data still includes it. On 8 of its 150 entries the older copy still carries a sentence already withdrawn from the maintained summary, which other planned work will bring into line. Everything that reads the older copy will be checked first; unless that check finds something that still needs it, the older copy will then be removed from the 15 records, with its earlier text kept in each record's correction notes. The maintained summaries, and every hazard value and risk level, stay as they are.
- D-1040
- State: queued · date filed: 2026-10-15
- Planned (counted 2026-10-15 in our records): 94 material and product records refer, in 167 places, to 45 compound record numbers that do not exist in our data. 40 of the 45 are numbers of alternative-name entries that were removed from our compound records, and our archive of that removal names the compound each stood for; our relationship listings silently skip these references. Each reference in a data link will be checked against its own wording and that archive, then pointed at the right compound, removed, or left as it is where the two do not agree; references written inside notes are left for a separate review and change history is not rewritten. The program that builds the listings will also count and report the references it skips.
- D-1042
- State: queued · date filed: 2026-10-20
- Not yet summarised for publication.
- D-846
- State: partial · date filed: 2026-09-21
- In progress: two pictures on our analytics page are not drawn from real data: a network of 15 items with invented links, under figures describing 2,300, and a flow diagram built from invented numbers. Every value will name a dated source or not be shown; the network will use the real data, and the flow diagram will too or be removed.
- D-849
- State: queued · date filed: 2026-09-21
- Planned: counts of our records are typed by hand into many published files, and nothing keeps them in step with the data. Each typed count will come from one measured source or be declared an exception, a count that disagrees will fail the build, and figures with no known source will be sourced or removed.
- D-858
- State: queued · date filed: 2026-09-21
- Planned: for every data source we use, record the address of its terms of use, an archived copy of those terms, and what they say about resale.
- D-913
- State: partial · date filed: 2026-09-29
- Planned: keep each substance's EC number in one place, and record which chemical groups a substance belongs to, with a source for each membership.
- D-932
- State: queued · date filed: 2026-09-29
- Planned: replace 1,615 unsourced "Not assigned" statements about ACGIH workplace limits, settle 33 ACGIH figures, and fix ACGIH values copied into compound and material text. Each value will be either marked "not checked" or checked under a licence.
- D-945
- State: queued · date filed: 2026-09-29
- Planned: we record what we do not know about each compound (for example, no published LC50), but the API's main compound response and the compound pages do not show it. A short curated "what we don't know" list will be added to both, leaving out bookkeeping entries and correction history.
- D-954
- State: queued · date filed: 2026-09-29
- Planned: the public description of our manifest-verification tool on the Python Package Index still names a product we have retired. A new version will drop that name from the description; uploading it is the operator's step.
- D-964
- State: queued · date filed: 2026-09-29
- Planned: add a booking link to the sites, and list Google as a service provider wherever a booking collects names or e-mail addresses.
- D-995
- State: queued · date filed: 2026-10-08
- Planned (counted 2026-10-13 in our records): our compound records carry a "not checked" wording (all 1,887 compound files) and a source and date on values, but no page on the site explains these labels to a reader; the only pages containing the words use them in other senses (accessibility statements, the corrections page) or in code comments. We will publish one plain page, linked from every compound page, that defines exactly the labels the pages show, each in one sentence with an example from a real record. A label that is not shown yet, such as "manufacturer-stated", which is on no record today, will not be defined until it is shown. No value changes.
- D-996
- State: queued · date filed: 2026-10-08
- Planned (checked 2026-10-13 in our records): our code has rejected API keys passed in the web address since 2026-04-15 (we have not tested the live API for this), but no page on the site records that removal, and we publish no versioning or deprecation policy saying how changes are classed, how much notice is given or how a retirement is signalled. The changelog's description says it covers deprecations in general. We will publish one policy page, listing past removals and stating the notice period, and generate the response headers, the changelog and the page from one list so that a retired signal cannot outlive its feature. Code left over from the removed feature will be taken out.
- D-997
- State: queued · date filed: 2026-10-08
- Planned (checked 2026-10-13 in our records): Article II of our Regulator Bill of Rights says a regulator sees the same values as anyone else, and that no paid mode changes a value, softens a finding or hides a disagreement between agencies. It says nothing about whether payment, plan or account status could affect how a report or correction is reviewed or in what order. We have not built a route in our API for receiving reports or review requests; readers' reports go through a general feedback form today. We will add one sentence to Article II, worded as a property of the system rather than a promise about any person, and only together with a test that checks on every release that no review code reads billing or plan fields.
- D-999
- State: queued · date filed: 2026-10-08
- Planned (checked 2026-10-13 in our records): we have not yet measured how much upkeep our records need each year, and we have no published figure for it, so any list of substances to add next would rest on a guess. We will first measure it from our own change history, counting how often records of each type change per year, and state the method and the window used. Then we will publish a short candidate list ordered by hand, each candidate with a written reason, a list of the inputs still missing and the checks it passed, plus a log of requests that is kept and answered but not used as a score. No value on any existing record changes.
- D-1000
- State: queued · date filed: 2026-10-08
- Planned (counted 2026-10-13 in our records): the site feedback widget, loaded on four of our pages, sends readers' reports, with an optional email address, to a Google form. On three of those pages the integrity check that the browser applies to the widget file does not match the file in our repository, so the widget may not run there and the "report a gap" route may do nothing. We will move the widget to a feedback endpoint on our own API, using the storage and rate-limit approach already in use and already named in our privacy policies, with no new vendor. The integrity check will be recomputed from the file we ship, and both privacy policies will say what the widget collects, where it is kept and for how long.
- D-1008
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): 22 claims in our public pages and files may be relied on by readers, such as how long email logs are kept, that everything we provide is exportable with each value's source and date, that ALETHEIA stays current, a pricing line reading "Unlimited*" and a statement about prior partners. Our own commitments check lists 241 published promises: 201 pass, 35 are not yet measured and 5 are marked false, including the export sentence. Typed figures also differ from the records: the ALETHEIA home page in our repository states 185 regulatory agencies where counting our records gives 193 (a count of agency names as written, so spelling variants count separately), and 365 California Prop 65 listings where our records give 386. Each claim will be checked against a measurement or a source, then kept with its measurement and date, narrowed to what is measured, or withdrawn.
- D-1013
- State: queued · date filed: 2026-10-12
- Planned (counted 2026-10-12 in our records): some sources we cite may not be checkable or current. Our compound records cite 3,043 distinct web addresses on three hosts (PubChem, PubMed Central and the WHO document archive), some of which may block our check that a source is still current; 73 saved source files are missing from our archive's index, including ten search captures from EPA's CompTox tool; and the safety site's copy of the data-gap lists differs from the API's on 10 compound records. The API's source entries may also leave out the label that tells a reader a link is only a finding aid or a model estimate; if so, 2,357 "where to look" entries and 779 model-estimate entries are affected. Each citation will be re-read at its source, then re-sourced, qualified, withdrawn to "not checked" where it cannot be supported, or reported.
- D-1022
- State: queued · date filed: 2026-10-14
- Planned (checked 2026-10-14 in our records): an earlier review found that two article identifiers (DOIs), cited on two compound records and one product record, are registered to papers other than the ones cited. Once we have archived each identifier's registered record and confirmed this, the wrong identifier will be removed and the citation text kept as a pointer; finding the right identifier is later work. We will also archive the registered records behind 20 identifiers that were corrected, kept or added in July (18 of them have no entry in our archive of saved sources, and the copies we hold for the other 2 will be checked), and a 2021 US House of Representatives staff report on baby food that some of our figures rely on.
- D-1024
- State: queued · date filed: 2026-10-14
- Planned (counted 2026-10-14 in our records): on 17 compound records, the hazard-code block (GHS classification) that companies notified to the European Chemicals Agency disagrees with something else on the same record. On one record the block is labelled as notified while its source text names the EU's official harmonised list; on 8, a note still says the block was withdrawn, although it was served again the same day; on 8 more, notes on field names no record uses say no hazard data exists, beside a block built from data that does. For each record the wrong statement will be decided from the record's own history and archived sources, then corrected where that evidence settles it; a record where it does not is left as it is. No hazard code changes and no block is withdrawn.
- D-1027
- State: queued · date filed: 2026-10-14
- Planned (counted 2026-10-14 in our records): each correction note on a record should name the written decision that approved it and the piece of work it belongs to. 985 notes (51 on materials, 934 on compounds) give the decision only as a shorthand reference or name a file that never existed, and 1,148 more (977 on materials, 171 on compounds) name the decision but not the work. The true decision and work will be found from each record's own history and filled in only where they can be shown; notes where they cannot stay as they are and are listed. These notes are part of our published data files; no value or page text changes.
- D-1035
- State: queued · date filed: 2026-10-15
- Planned (counted 2026-10-15 in our records): when a source was attached to a dose value without changing the value, one piece of work recorded it as a correction and another proposed recording nothing. 220 notes on 76 compound records record only an added source, with the value and its notes unchanged, so calling them corrections overstates what happened. One rule will be written down, and those 220 notes removed or relabelled under it; no value changes, and the count of data gaps shown for those records is to stay the same.
- D-1038
- State: queued · date filed: 2026-10-15
- Planned (counted 2026-10-15 in our records): our check of whether each correction rests on a document a reader can open counts some withdrawn values as if they had been replaced with new ones. In 32 notes recording a withdrawn intake or reference-dose figure, the value shown is already empty or "not established" and the withdrawn figure is kept in the note; 1,583 notes on 1,430 records, and 55 text entries, say a value was not checked but not in the form the check recognises. Each kind will be reviewed and ruled on, and notes ruled to be withdrawals written in an older form will be restated in the recognised form; no value shown changes, and nothing is restated just to change a count.
- D-1039
- State: queued · date filed: 2026-10-15
- Planned (counted 2026-10-15 in our records): our archive of saved source documents marks 472 saved copies as not containing the document, for example because the copy is a bot-check page or a "not found" page. 61 of them are copies we saved by hand, which are handled separately; the other 411 have not been checked against that meaning, and a first automatic pass with our capture tool's own checks found 217 that are neither a bot-check page nor a "not found" page. Each of the 411 will be read and its mark corrected where it is wrong, because a copy wrongly marked is left out of our counts of archived sources. No page or served value changes.
- D-1041
- State: queued · date filed: 2026-10-15
- Planned (checked 2026-10-15 in our records): the product record for nail salon manicure, pedicure and gel nail services lists HEMA and TEGDMA under the heading "Photoinitiators", but our own records for both substances describe them as monomers, a different kind of ingredient; no other record in our data does this. The entry's ingredient name, role and stated 1-5% concentration, none of which cites a source, will be reviewed together: each will be corrected to match our records, withdrawn, or kept only if a saved source supports it. No risk band is calculated from this entry, so no band is expected to change.
- D-1043
- State: queued · date filed: 2026-10-21
- Not yet summarised for publication.