Summary

  • The Bylaws set two separate walls against the Board overruling advice — no less than 60% of the Board for GAC Consensus Advice, more than two-thirds of the Board for a GNSO Supermajority recommendation — and grade the duty by the adviser's own internal margin, not by the stakes of the advice.
  • Article 2, Section 2.1 then defines a vote of the Board as the vote of the Directors present at a quorate meeting, and neither advice provision invokes the Bylaws' own 'of all Directors' form. The printed fraction is a full-attendance ceiling, not a floor.
  • A reader can recompute the fraction. The vote needs the attendance, and the attendance sits in the Board's record of that meeting.

Article 12, Section 12.2(a)(x) begins with a duty that binds on Governmental Advisory Committee advice as an institution: the Board takes it duly into account, and where it acts inconsistently it must inform the GAC and state the reasons why it declined to follow. Then the same sentence adds a narrower class. Advice approved by a full GAC consensus, understood as the practice of adopting decisions by general agreement in the absence of any formal objection — GAC Consensus Advice — "may only be rejected by a vote of no less than 60% of the Board", after which the two sides try in good faith to find a mutually acceptable solution. The GAC states which of its advice carries the label. ICANN quotes the provision verbatim in its approved resolutions of 26 January 2025.

So the reason-giving duty attaches to GAC advice as a category, while the vote wall attaches only to the slice the GAC has decided to call consensus. A communiqué that three governments contested attracts a duty and no threshold. A communiqué adopted with no formal objection attracts a supermajority requirement. The adviser grades its own advice, and the Board's obligation rises or falls with the grade.

Watch the vocabulary, because ICANN's own explainers blur it. A New gTLD Program FAQ describes GAC Consensus Advice as representing "the collective view of governments participating in the GAC, rather than the position of an individual government" (what is GAC Consensus Advice). That is a description of a defined term, not the term. The Bylaws do not qualify the duty with the word collective; the operative qualifier is the consensus label, and the GAC applies it.

The same design runs on the policy side. Annex A, Section 9(a) governs GNSO Council recommendations with two thresholds rather than one. A recommendation carried by a GNSO Supermajority Vote is adopted unless the Board finds, "by a vote of more than two-thirds (2/3) of the Board", that it is not in the best interests of the ICANN community or ICANN. A recommendation where the Council fell short of a supermajority needs only "a majority vote of the Board" to be set aside. Section 9(b) still demands a Board Statement giving reasons, and Section 9(d) applies the same more-than-two-thirds test to a Supplemental Recommendation — the annex text as it circulates in the GNSO Council list archive.

Put the two provisions side by side and the pattern is unmistakable. ICANN did not build one override rule for community advice and scale it by consequence. It built a graduated system in which the adviser's internal cohesion is the variable: the closer the GAC or the Council comes to speaking as one voice, the more Board votes it takes to answer back.

Then comes the base, and the base is the room. Article 2, Section 2.1 of the Bylaws, on the page marked "As amended 3 July 2026", defines a vote of the Board as the vote "of only those Directors present at the meeting where a quorum is present unless otherwise specifically provided in these Bylaws by reference to 'of all Directors'" (Bylaws). The carve-out is the tell. This instrument knows how to pin a denominator to the whole body, and it does so elsewhere: Section 3.6, on the same page, uses the majority-vote-of-all-Directors form for its own matters. Neither Section 12.2(a)(x) nor Annex A, Section 9(a) invokes that form.

Which leaves composition. Sixteen voting Directors plus four non-voting Liaisons, with only Directors counted toward quorums and the validity of votes — the Article 7 language reproduced in ICANN's proposed Bylaws amendments of 10 April 2023. The consolidated Bylaws page did not render that far for me on two attempts, so the corroboration deserves its label: a Namecheap pre-hearing brief from the Independent Review Process, filed by a party with every reason to quote the composition accurately, uses the same words.

Now the arithmetic, which is mine and not the instrument's. Against sixteen Directors at full attendance, sixty percent is 9.6, so at least ten votes; more than two-thirds of sixteen is more than 10.67, so at least eleven. Both figures are ceilings on what the printed text demands, not floors. Because Section 2.1 measures the fraction against Directors present, fewer people in the room means fewer absolute votes clear the same wall. And the quorum number was not read verbatim in this review: Article 7, Section 7.14 sets it, and a quorum at a majority of the Directors then in office is my inference rather than a quotation.

What follows is a checkability problem, and it is structural rather than scandalous. To test whether the Board actually cleared either wall at a real meeting you need three items recorded in three different places: the adviser's classification of its own advice, the number of Directors present and voting, and the tally. The advisory bodies publish the first, and the machinery is written out in ICANN's GAC Advice Consideration Process and still taught to incoming members in the GAC's operational-basics material, which in May 2026 restates the sixty-percent rejection bar. The second and third sit in the Board's own record of that meeting.

So the Bylaws print the fraction precisely and print the base loosely, which leaves the fraction unverifiable from the instrument alone. A reader can recompute the wall. A reader cannot recompute the vote. If you set out to draft a threshold that reads as a hard protection and behaves as a movable one, this is the shape you would arrive at — and I hold no document showing that anyone built it with that intention. The drafting contrast with Section 3.6 makes deliberate choice the more plausible reading of the text, and plausible is as far as the text carries me.

Three boundaries, stated plainly. The consolidated Bylaws page truncated before Articles 7 and 12 both times I retrieved it, so the operative Article 12 sentence comes from ICANN's own block quotation in the January 2025 resolutions rather than from my reading of the article itself. Section 12.2(a)(xi), which ICANN's handbook names alongside (x) as part of the same procedure, was not obtained in readable form here; I describe the process as the handbook describes it and claim nothing about what (xi) contains.

And no attendance record or vote tally for any specific advice decision entered this review, which is why nothing above characterises any Director's conduct.

What would change the reading is narrow and dull and decisive: one published tally sheet per advice decision, carrying the advice classification, the Directors present and the votes cast. That would turn a structural observation into an auditable one and let anyone see whether the gap between the printed fraction and its absolute requirement has ever mattered. Until it exists, the accurate statement is sharper than either side of this argument usually wants. ICANN's advice thresholds are graded by the adviser and counted by the room, and the second half of that sentence does more work than the first.

Institutional reference: Internet Corporation for Assigned Names and Numbers.