Rule · MAJOR

COMMON.LEI_CHECK_DIGITS

Legal Entity Identifier format and check digits (ISO 17442, ISO/IEC 7064 MOD 97-10)

How the status is set

• Missing: no section that reads LEIs on its own (below) holds an LEI-shaped code (18 upper-case letters or digits and two check digits) as written, and no section issuer, offeror or trading_platform_operator has text other than a placeholder. A malformed code, or an LEI-shaped code elsewhere that fails its check digits (below), is named in the finding but does not change this status.
• Review: an LEI-shaped code that is read fails the ISO/IEC 7064 MOD 97-10 check digits; an LEI-like code that is read is malformed; or no section that reads LEIs on its own holds an LEI-shaped code as written while a party section has text, since point 5 allows another identifier required pursuant to applicable national law in place of the LEI. A section reads LEIs on its own if it is issuer, offeror or trading_platform_operator, names an LEI label (LEI or LEIs, legal entity identifier or ID, Rechtsträgerkennung, in any case, also split by hyphens, dashes or line ends) or holds an LEI-shaped code with a letter. A draft with such a section, or whose file text names a label or holds such a code, is read in all its text: every section and, for a PDF, DOCX, Markdown, XHTML or HTML file, the text outside the sections (headings, tables, text boxes, headers and footers, text in no section). Each text is read as written and, where markup or a character reference changes it, as a Markdown or HTML renderer shows it (comments, tags, emphasis, code and link markup removed,   and other references replaced by their characters). There every LEI-shaped code, a 20-digit number included, has its check digits checked, and every LEI-like code counts, wherever it stands: one run of 16 to 24 letters and digits of any script that holds a digit or is in ASCII capitals, or a longer run with a digit that a glued LEI (up to 28 characters) or a glued word explains; groups joined by one to three spaces, hyphens, dashes or line ends, with at most one hyphen or dash and one line end between two groups, and by any number of invisible characters (zero-width spaces, bidirectional marks and other format characters), that together hold at least 16 and hold a digit or all have one length; or two pieces with letters and digits of a code that a table cell wraps from a line that names an LEI label, or the line before one, onto the next line. A word of letters only with a lower-case letter or a letter outside ASCII joins only beside a group of its length with a digit. An LEI-like code is malformed unless it is one run that is LEI-shaped as written. So a code is malformed if it has 16 to 19 or 21 to 24 characters, is in lower case, has a letter where a check digit belongs or a character outside ASCII, is glued to LEI or to a word, is split into groups, or is wrapped in a table cell.
• Pass: a section that reads LEIs on its own holds an LEI-shaped code as written, every LEI-shaped code that is read, as written or as shown, has valid check digits and no LEI-like code is malformed. No registry is asked.

Legal basis

Annex I, Part A, point 5, Part B, point 5, Part C, point 5, Annex II, Part A, point 5, Annex III, Part A, point 5 MiCAR

• https://eur-lex.europa.eu/eli/reg/2023/1114/oj/eng

Checked on 2026-10-27 against the Official Journal text in English and German: point 5 of Annex I Parts A, B and C and of Annex II and III Part A names the legal entity identifier or another identifier required pursuant to applicable national law, and Art. 2(1)(b) and (c) of Implementing Regulation (EU) 2024/2984 ask for the LEI where available.

Synthetic examples

Each example is linted when this page is generated. The build fails if a result differs.

{
  "title": "Synthetic rule example",
  "type": "other",
  "sections": {
    "summary": "Synthetic summary of a fictional token."
  }
}

{
  "title": "Synthetic rule example",
  "type": "other",
  "sections": {
    "issuer": "Synthetic Issuer AG."
  }
}

{
  "title": "Synthetic rule example",
  "type": "other",
  "sections": {
    "issuer": "Synthetic Issuer AG. LEI: 9999SYNTH0000ZZQ0167."
  }
}

{
  "title": "Synthetic rule example",
  "type": "other",
  "sections": {
    "issuer": "Synthetic Issuer AG, a fictional company. LEI: 9999SYNTH0000ZZQ0166.",
    "offeror": "Synthetic Offeror GmbH. LEI: 9999SYNTH0000ZZQ026."
  }
}

{
  "title": "Synthetic rule example",
  "type": "other",
  "sections": {
    "issuer": "Synthetic Issuer AG, a fictional company. LEI: 9999SYNTH0000ZZQ0166.",
    "offer_or_admission": "Synthetic Admission GmbH (9999SYNTH0000ZZQ026) seeks admission to trading."
  }
}

{
  "title": "Synthetic rule example",
  "type": "other",
  "sections": {
    "issuer": "Synthetic Issuer AG, a fictional company. LEI: 9999SYNTH0000ZZQ0166."
  }
}

Limitations

• Reads every section of the draft; findings are anchored to the issuer section.
• Works offline on the characters of each code. It does not confirm that an LEI is registered, current or held by the entity the draft names.
• Status and severity are fixed by the rule. A draft cannot waive or downgrade them.
• Does not recognise the other identifiers that point 5 allows in place of the LEI, so a draft without any LEI goes to review once a party section has text. Once the draft gives one LEI, a party section without an LEI of its own does not change the result.
• Reads the sections and, for a document file, the text its loader extracts. Text the loaders do not extract is not seen: in a DOCX, footnotes, endnotes, comments, deleted text, field codes and text in charts, SmartArt or images; in a PDF, text that is not extracted (images, scans, annotations, form fields); in XHTML or HTML, the head, scripts, styles and attribute values, and CDATA sections, which the HTML parser drops (a code inside <![CDATA[…]]>, or the part of a code inside one, is not read); in a JSON draft, anything outside its sections, such as the title. An entity that an XHTML or HTML document declares in its own DTD is not expanded, so a code given through one is not read. Hidden text is read as if it were shown: a DOCX run marked hidden and XHTML or HTML text that CSS hides (display:none), so a hidden fragment that completes a code the document shows malformed makes it read as a valid LEI, and the rule passes. Nor is a malformed code that is no LEI-like code in the text as written or as shown: one run of fewer than 16 letters and digits, or of more than 24 (28 for one with a digit after a glued LEI) that no glued word explains, as when a code has lost or gained five or more characters or has digits glued to it; groups that together hold fewer than 16; a run of letters only with a lower-case letter or a letter outside ASCII, which reads as a word; groups split by any other character, such as a full stop, a slash, a comma, a bullet or an underscore that a renderer shows, by more than three characters other than invisible ones, or by two hyphens or dashes or two line ends (a blank line); a code split into groups where such a word stands beside no group of its length with a digit; groups in ASCII capitals without a digit and of different lengths; or a code that a table cell wraps where a piece holds no letter or no digit or is split into groups, where neither line names an LEI label (as in a PDF party table whose LEI label stands only in its header row, above the row in which the code wraps), where other lines stand between its pieces, or where the code wraps over more than two lines. Markup inside text that a renderer does not show, such as an HTML comment or a link target, is read as written only. Every LEI-like code in a draft that reads LEIs counts, wherever it stands, so an IBAN, a long number in groups, a placeholder such as XXXXXXXXXXXXXXXXXXXX, a reference number of 16 to 24 letters and digits, an address or a heading in capitals with a digit in it, a word of 16 to 24 ASCII capitals in a heading or in text, numbers in neighbouring table cells that join, two codes with letters and digits on a line that names an LEI label and the next line, or a valid LEI glued to LEI or to a word or split into groups (also by markup, as written) gives Review too, which fails safe. A code that the file text and a section quote with different neighbours is named twice.

Disagree with this rule, or missing one? Propose a rule change.