Summary

  • Consensus advice from the Governmental Advisory Committee (GAC) placed Amazon’s three applications behind a powerful institutional barrier, but it did not itself reject them. The New gTLD Program Committee (NGPC), acting under authority delegated by the ICANN Board, made the 14 May 2014 “Will Not Proceed” decision. The 2017 Independent Review Process (IRP) majority later found the method of that decision inconsistent with ICANN’s governing obligations and recommended prompt re-evaluation; it did not award the strings, compel Registry Agreements or order a root-zone change.
  • The programme’s Independent Objector Community Objection was a different track. In a consolidated expert determination, Amazon prevailed because the Objector did not prove substantial community opposition or likely material detriment, even though the expert found a strong association between the strings and the invoked Amazon community. That dismissal did not cancel GAC advice or bind the Board’s response to it.
  • ACTO and its member states gained meaningful leverage and later obtained specified contractual rights, but not a general consent veto or joint technical control. The final Specification 11 language created nine non-commercial names and up to 1,500 permanent blocks in each of the three TLDs, a cultural-term non-use obligation and exclusive ACTO standing for disputes about that commitment. The agreements were then assigned to Amazon Registry Services, Inc.; public readiness reports dated 28 May 2020 confirmed successful Pre-Delegation Testing; and IANA’s 2 June reports documented the separate delegation review.
  • The result is therefore a chain of necessary but individually limited public evidence institutional acts. Review corrected a decision process; the Board reopened the programme route; negotiation supplied terms; publication and contracting converted selected terms into obligations; assignment placed them on the eventual operator; and readiness and root-zone processing made delegation executable. Treating any one stage as the source of the entire outcome obscures participation, decision power and remedy.

Four documents, four different powers

Place four primary records side by side and the authority chain becomes easier to see.

The first is the 11 July 2017 Independent Review Process Final Declaration. Its majority concluded that the Board, acting through the NGPC, had not exercised the independent, objective and reasoned judgement required by ICANN’s governing commitments when it accepted governmental advice against the applications. The declaration recommended prompt re-evaluation. It did not say that .AMAZON, the Japanese .アマゾン or the Chinese .亚马逊 had to be approved or delegated.

The second is the 15 May 2019 Board resolution. The Board found Amazon’s proposed safeguards acceptable, identified no public-policy reason why the applications should not proceed, and directed continued processing and publication of the proposed commitments for comment. It did not execute a Registry Agreement or instruct the root-zone system to add the strings.

The third is the 19 December 2019 .AMAZON Registry Agreement, read with the parallel Japanese-IDN agreement and Chinese-IDN agreement. Those contracts designated a registry operator, made the ACTO Commitment part of Specification 11 and expressly preserved the requirements and necessary approvals for delegation and root-zone entry. Signature created enforceable obligations; it did not itself alter the root.

The fourth is the delegation dossier. Public String Delegation Readiness reports for .AMAZON, the Japanese variant and the Chinese variant, all dated 28 May 2020, record confirmed readiness and successful Pre-Delegation Testing. IANA then published delegation reports for .AMAZON, xn--cckwcxetd and xn--jlq480n2rg on 2 June. Those reports document the proposed manager, consistency with the contracted party, contacts, technical conformance and completion of required processing.

The records are connected, but none substitutes for another. The IRP could not sign a contract. The Board resolution could not complete technical testing. A Registry Agreement could not insert a string into the root merely by naming an operator. The IANA process did not reopen the geopolitical merits. The useful question is therefore not simply who “won”, but which institution could make the next legally or operationally necessary act.

Three applications, two corporate roles and no entitlement to a string

The case concerned three applications submitted in the 2012 New gTLD round. The ASCII application was .AMAZON, application 1-1315-58086. The Japanese internationalised string was .アマゾン, application 1-1318-83995, represented in the DNS as xn--cckwcxetd. The Chinese string was .亚马逊, application 1-1318-5591, represented as xn--jlq480n2rg.

The Chinese identifier requires a small but important documentary caution. One operative line in the 14 May 2014 NGPC resolutions transposes a digit and prints 1-1318-5581. The same Board record’s rationale, the application material and the IANA delegation record support 1-1318-5591. The supported identifier is therefore 1-1318-5591.

The original applicant and the party that signed the 2019 agreements was Amazon EU S.à r.l. The eventual operator named in the readiness and delegation records was Amazon Registry Services, Inc. That distinction matters because an application, an executed agreement, an assigned agreement and an operating sponsorship are different institutional entities. An applicant can seek evaluation, answer objections, invoke accountability procedures, negotiate and offer commitments. It cannot approve its own application or place a preferred string in the root. A contracting party can accept obligations and prepare for delegation.

It must still satisfy the programme’s operational conditions. An assignee can take contractual rights and liabilities, but assignment alone does not prove technical readiness or complete a root-zone change.

The .AMAZON application-status record confirms the identifier, original applicant, evaluation history, objection result and later delegated status. It does not establish that filing created a property right to the string. The New gTLD process was built around staged eligibility and institutional decisions. An application was a request to enter that process, not a conveyance of the namespace.

This point is easy to lose because the later agreements classify the strings as dot-brand TLDs under Specification 13. Dot-brand status describes a contractual operating model; it does not retrospectively turn the 2012 filings into vested rights. Nor did it eliminate the separate Specification 11 commitments negotiated for the Amazonia region. The BAMC recommendation on Colombia’s 2019 reconsideration request, later adopted by the Board, treated the two contractual layers as capable of coexisting.

The corporate succession makes the same point from another direction. ICANN’s completed assignment register records that all three agreements moved from Amazon EU S.à r.l. to Amazon Registry Services, Inc. on 14 February 2020. Authority followed the recognised agreement and operator role, not a general claim of ownership over the string.

Governmental warning carried force without deciding the case

The first public-policy interventions came from Amazon Basin governments. Brazil and Peru issued Early Warnings, with support recorded from Bolivia, Ecuador and Guyana. The 14 May 2014 New gTLD Program Committee resolutions reconstruct that history and the later consensus advice from the Governmental Advisory Committee at Durban.

An Early Warning mattered because it gave the applicant formal notice that one or more governments saw a public-policy problem and created a channel for response, withdrawal or negotiation. It did not reject the applications. The governments issuing or supporting the warnings did not control the application-status database, execute registry contracts or manage the root zone. Their effective power was to raise the political and procedural cost of proceeding without an answer.

The GAC’s consensus advice was stronger. Under section 3.1 of the Applicant Guidebook, consensus advice that an application should not proceed carried a strong presumption for the Board. The Durban advice therefore changed the institutional burden. It made approval less likely and required the Board to confront the governmental position through the Guidebook’s consultation and explanation framework. Yet the presumption did not transfer final programme authority to the GAC. The advice needed a Board or delegated Board act before application status changed.

That division is neither semantic nor ceremonial. Advisory force can be substantial even when advice is not legally self-executing. A Board may know that departure from GAC consensus will require consultation, reasons and acceptance of institutional conflict. Governments may therefore achieve much of the practical effect of a veto without possessing a formal one. But accountability still depends on identifying the actor that converts advice into a decision. In 2014 that actor was the New gTLD Program Committee, not the GAC itself.

The difference also defines the available remedy. An applicant dissatisfied with GAC advice could submit information, lobby governments, seek consultation or challenge the Board’s treatment of the advice. It could not use the accountability system to command the GAC to approve the string. The central reviewable act was the Board-side decision that gave the advice programme effect. That is why the later IRP focused on whether ICANN’s decision-maker had exercised the independent judgement required by the Bylaws, rather than asking whether governments were entitled to prefer protection of regional cultural and geographic interests.

The same distinction limits claims about governmental consent. Individual states could warn, negotiate, comment and seek reconsideration. ACTO could aggregate regional positions. The GAC could issue consensus advice. None of those acts, standing alone, was a registry agreement, a pre-delegation approval or an IANA root-zone instruction. The record supports governmental leverage and privileged advisory standing. It does not support a general state consent right over every subsequent programme stage.

The community objection was a separate test

The Independent Objector’s Community Objections formed a separate adjudicative track from GAC advice. The objections were consolidated and decided in a 3 February 2014 expert determination covering .AMAZON and both internationalised variants.

The determination is more informative than the application-status label alone. The expert first upheld Amazon’s challenge to the Independent Objector’s independence, treating the perceived neutrality of the objection system as material. He nevertheless addressed the merits. Under the Applicant Guidebook, a Community Objection could prevail only if all four cumulative requirements were met: a clearly delineated community, substantial opposition within that community, a strong association between the community and the applied-for string, and a likelihood of material detriment to a significant portion of the community.

The Objector proved only part of that case. The expert found a strong association between the strings and the invoked Amazon community. He did not find substantial opposition within that community and did not find the required likelihood of material detriment. Because those failures were sufficient to defeat the objections, he did not need to reach a conclusive answer on clear delineation. The objections were dismissed and Amazon prevailed in that track.

That result did not decide the GAC-advice track. The expert did not order the NGPC to approve the applications, cancel the Durban consensus advice or compel ICANN to contract. Community Objection and GAC advice had different decision-makers, standards and consequences. Amazon could win one without controlling the other. Keeping the tracks separate is essential because the later accountability dispute concerned how the Board treated governmental advice, not whether the Independent Objector had met the Guidebook’s four-part test.

The 2014 “Will Not Proceed” decision belonged to the NGPC

On 14 May 2014, the New gTLD Program Committee accepted the GAC advice and decided that the three applications should not proceed. The approved resolutions identified the Early Warnings, the Durban consensus advice and the Guidebook’s strong-presumption standard. The dispositive act was an NGPC resolution placing the applications in “Will Not Proceed” status.

The NGPC was exercising authority delegated by the ICANN Board for the New gTLD programme. That institutional location matters. The GAC supplied advice. The NGPC supplied the programme decision. Describing the result simply as a governmental rejection would erase the actor whose reasoning and compliance with ICANN’s governing instruments could be challenged.

Amazon filed Reconsideration Request 14-27 on 30 May 2014. The request advanced the applicant’s claims against the decision and sought reversal or further consideration. Those claims were party submissions, not findings. Reconsideration was not a fresh merits appeal in which every disagreement with the public-policy outcome would automatically be reheard.

The NGPC’s 8 September 2014 resolution denying the request illustrates the mechanism’s narrower character. Reconsideration examined whether the challenged action rested on materially false or inaccurate information or conflicted with governing policies and procedures. Access allowed Amazon to make a formal challenge and required an institutional response. It did not guarantee re-adjudication of the geopolitical merits or continuation of the applications.

This was the first sharp demonstration of the difference between review access and remedy. Amazon had standing to seek reconsideration and received a decision, but not the relief it wanted. Governments could point to a Board-side act implementing their consensus advice, but they did not acquire ownership of the programme outcome. The status rested on an ICANN decision that could later be tested under another accountability mechanism.

The IRP majority corrected the method, not the destination

Amazon then invoked the Independent Review Process. The resulting Final Declaration of 11 July 2017 was the most consequential accountability intervention in the case, but its consequence is often overstated. The panel majority did not decide that Amazon had a substantive entitlement to the strings. It decided that ICANN’s decision-making had not met the standard imposed by its governing commitments.

The majority’s first concern was independent judgement. GAC consensus advice was entitled to substantial procedural weight, but the Board could not treat it as conclusive merely because governments had reached consensus. The NGPC still had to evaluate the applications and the public-policy concerns in an independent, objective and reasoned manner. On the majority’s reading, the 2014 record showed effectively complete deference rather than a discernible merits assessment by the body that held decision power.

The majority also focused on reasons. At the time, the GAC’s procedures did not require it to provide a full merits explanation with its consensus advice. That did not release the Board from its own obligation to explain why the advice justified the result it adopted. The accountability defect therefore lay in the Board-side conversion of advice into a dispositive decision without the independent assessment required by ICANN’s governing framework.

A second majority conclusion concerned procedural fairness. The majority considered the GAC’s failure to give Amazon an opportunity to present information to it inconsistent with minimum fairness and treated that failure as diminishing the presumption otherwise attaching to the advice. That proposition must be presented as a conclusion of the majority in a divided declaration. The dissent disputed both the procedural-fairness theory and the extent to which IRP review could reach the GAC-related process.

The disagreement marks an institutional boundary. A review panel can require the Board to comply with the Bylaws without becoming the primary allocator of a TLD. It can examine whether the Board’s reliance on another ICANN body was compatible with ICANN’s procedural commitments without claiming managerial authority over that body. The majority increased the practical demand for independent reason-giving; the dissent warned against expanding review beyond its proper institutional target.

The remedy preserved the separation. The panel declared that the challenged ICANN action was inconsistent with the Bylaws, treated Amazon as the prevailing party and recommended prompt re-evaluation followed by an independently reasoned decision. It did not order the Board to approve the applications, direct ICANN organisation to execute agreements or instruct IANA to delegate the strings. The declaration changed the accountability position and made it difficult to leave the 2014 decision untouched, but a new Board act was still required to change programme status.

That limit was not a technical deficiency in the remedy. The entity under review was the consistency of ICANN action with its governing instruments. It was not a root-zone change request, an executed Registry Agreement or a technical-readiness report. A declaration capable of correcting process without allocating the contested resource preserved the distinction between review and primary decision-making.

The counterfactual clarifies the declaration’s force. Without the IRP majority’s finding, “Will Not Proceed” could have remained the operative programme result. With the declaration alone, however, the applications still could not advance. The victory was neither empty nor self-executing: it removed the institutional legitimacy of leaving the earlier decision in place without re-evaluation while preserving the Board’s substantive choice.

Re-evaluation required reasons, and the GAC changed its advice

The Board’s response confirms that reading. In its 29 October 2017 resolutions, it treated the IRP recommendation as non-binding but resolved to undertake the requested re-evaluation. It asked the GAC for merits-based public-policy reasons and for new or additional information relevant to the applications.

That step did two things. First, it reasserted that the Board, not the IRP panel, would make the programme decision. Second, it acknowledged that independent judgement required more than repeating the existence of consensus advice. The Board needed an evidentiary and reasoned basis on which to accept, reject or modify the effect of that advice.

The GAC’s later position did not simply reproduce Durban. In the Abu Dhabi communiqué, the GAC advised ICANN to continue facilitating negotiations between ACTO member states and Amazon towards a mutually acceptable solution that would allow use of the strings. The institutional signal had shifted. The earlier advice said the applications should not proceed. The later advice contemplated progression if negotiations produced an acceptable arrangement.

That shift did not mean the Amazon Basin governments had abandoned their concerns, nor did it establish that any proposal from Amazon would suffice. It changed the procedural route from presumptive non-progression to facilitated bargaining. The entity of power was no longer only application status; it was also the content of possible safeguards and the terms on which the Board might permit processing to resume.

On 25 October 2018, the Board removed the “Will Not Proceed” status and directed ICANN organisation to resume processing. It treated the Abu Dhabi advice as superseding the Durban advice for the next stage. The resolution did not delegate the strings. It reopened a route through the programme and set the conditions for further negotiation and evaluation.

This is another point at which compressed narratives fail. Saying that the Board “approved .AMAZON” in October 2018 would skip the unresolved safeguards, the possibility of further Board intervention, contracting, assignment, technical readiness and IANA processing. Saying that governments “withdrew their objection” would understate the continued bargaining over cultural terms and use rights. The accurate institutional description is narrower: the Board removed a policy status that had prevented further processing and converted the revised GAC advice into a negotiation-centred pathway.

ACTO could negotiate hard, but the Board retained the decision

ACTO and its member states now had a structured channel through which to seek protections for names associated with the Amazonia region. That was meaningful power. A proposal unacceptable to the regional governments could generate renewed GAC concern, reconsideration requests, political pressure and delay. The Board had already learned from the IRP that it needed a reasoned record rather than conclusory deference. Amazon therefore had strong incentives to produce commitments that could be defended as responsive to the public-policy concerns.

Meaningful leverage, however, is not the same as programme control. ACTO could negotiate, define interests and press for consent, concurrent use or shared-governance arrangements. It could not amend the Applicant Guidebook, order the Board to restore “Will Not Proceed” status or execute a Registry Agreement for ICANN. The Board retained authority to decide whether any proposal was acceptable and whether processing should continue.

ACTO tested the October 2018 resumption through Reconsideration Request 18-10. The Board’s 16 January 2019 denial rejected the challenge without treating resumed processing as immediate delegation. The decision left room for continued negotiation and a later Board judgement. Reconsideration could test whether the October action complied with governing procedure; it did not create a new governmental consent requirement.

The Board then imposed a closing timetable. Its 10 March 2019 resolutions gave the parties a final negotiation period and made the retained decision options explicit. The Board could accept Amazon’s proposal, return it for further work or determine that the strings should not be delegated. The deadline increased pressure on both sides, but the menu of outcomes remained with the Board.

This structure distributed incentives precisely. Amazon needed safeguards sufficient to secure Board acceptance and reduce renewed governmental opposition. ACTO needed to translate broad regional claims into obligations that could be administered. The Board needed a record showing independent judgement rather than automatic deference to governments or automatic preference for the applicant. ICANN organisation needed language capable of publication, contract drafting and enforcement. Each actor could obstruct the pathway in a different way, but none could complete it alone.

The April 2019 proposal was a bridge, not yet a remedy

Amazon submitted its safeguard proposal on 17 April 2019. It offered the elements that later became central to the contractual settlement: non-use of terms with primary and well-recognised significance to the culture and heritage of the Amazonia region; nine domain names for non-commercial designation by ACTO and its eight member states; permanent blocking of up to 1,500 culturally significant terms; and access to the Public Interest Commitment dispute mechanism for failures relating to the ACTO Commitment.

The proposal mattered because it translated a political dispute into terms that could be tested for administrability. “Respect regional interests” would have been too indeterminate to monitor. A defined number of names, a ceiling on blocked terms, a class of protected cultural and heritage expressions and identified standing under a dispute procedure could be drafted into an agreement.

It was still an applicant proposal. Amazon could offer commitments, but could not declare them sufficient. ACTO’s response was politically important, but formal ACTO consent was not the sole programme test. The Board had to decide whether the package answered the public-policy problem well enough to permit continued processing. ICANN organisation then had to administer the proposed application change and contracting steps.

On 15 May 2019, the Board found the proposal acceptable and not inconsistent with the GAC’s advice. It concluded that there was no public-policy reason why the applications should not proceed, directed ICANN organisation to continue processing and required publication of the proposed commitments for a 30-day comment period. The Board made a substantive policy judgement, but did not skip the administrative and contractual steps needed to give that judgement operational effect.

The sequence is important. Negotiation generated the proposal. The Board assessed its sufficiency. ICANN organisation administered the Application Change Request and public comment. Contracting would determine the final binding language. A proposal could influence the Board without being enforceable; a Board endorsement could authorise continued processing without itself creating a private-law obligation; public comment could expose defects without deciding delegation.

Colombia challenged the May decision through Reconsideration Request 19-1. The 14 August 2019 BAMC recommendation examined claims that the Board had relied on incomplete or inaccurate information, failed to follow applicable procedure and inadequately considered alternative governance arrangements. It also addressed the relationship between Specification 13 brand status and the proposed Specification 11 commitments. The analysis rejected the proposition that dot-brand treatment would make the safeguards legally empty.

On 8 September 2019, the Board adopted the BAMC recommendation and denied the request. That final action matters: a committee recommendation did not itself dispose of Reconsideration Request 19-1. Colombia obtained a formal accountability response and tested the procedural record, but did not acquire power to draft the agreement, compel a governance seat or direct IANA. Denial also did not prove that every regional concern had been satisfied in practice. It meant that the challenge did not displace the May Board decision under the reconsideration standard.

Contract turned selected safeguards into enforceable obligations

ICANN’s 19 December 2019 status notice records the administrative bridge to contracting. Amazon submitted an Application Change Request on 5 June 2019. ICANN published the proposed commitments, completed public comment and decided that the applications were ready to proceed to contract. On 19 December, ICANN and Amazon EU S.à r.l. executed separate Registry Agreements for .AMAZON, xn--cckwcxetd and xn--jlq480n2rg.

The difference between the April proposal and the December agreements is the difference between an assurance and an enforceable institutional entity. A letter can show intent and support a Board decision. A Registry Agreement identifies the obligated party, incorporates commitments, supplies compliance machinery and connects breach to contractual remedies. The safeguard bargain acquired legal force because it was written into Specification 11 of each agreement, not because negotiators had announced convergence.

The three agreements are parallel, not inferential. Each contains a Section 4 ACTO Commitment tailored to its own TLD. The nine-name right and the maximum 1,500-name permanent block therefore apply in each agreement and each TLD; they are not a single pool silently shared across the ASCII and IDN strings. The same is true of the non-use obligation and PICDRP standing.

Specification 11 introduces the package by saying the operator is committed to “sharing the use and governance” of the TLD with ACTO. That phrase should not be erased, but neither should it be expanded beyond the operative clauses. The agreement does not create a joint registry operator, a second root-zone authority or an ACTO consent right over every operational decision. It implements the stated sharing through particular naming rights, reserved terms, standing and controlled-use arrangements.

The first substantive obligation is a non-use commitment. The operator must not use as domain names terms having primary and well-recognised significance to the culture and heritage specific to the Amazonia region. The language is bounded. It does not prohibit every expression associated with the region or grant ACTO general approval power over the registry’s naming policy. It creates an enforceable standard directed at a defined class of culturally significant terms.

The second element provides nine domain names in each TLD for ACTO to designate to itself and its eight member states for non-commercial purposes. Those names remain inside operator-controlled mechanics. The agreement’s naming policy provides that registration is made through the operator’s chosen registrar, in the name of the operator or a designated affiliate; the names are non-transferable and delegated to servers selected and controlled by the operator. ACTO and the states receive bounded use and designation rights, not independent control of the registry database, registrar channel or nameserver architecture.

The third element requires the permanent reservation, blocking all use, of up to 1,500 culturally significant domain names in each TLD. ACTO’s authorised representative could submit additions to the list until two years after the relevant agreement’s effective date. The operator could reject a request on specified grounds, had to communicate rejection in writing within 21 days, and ACTO could challenge that rejection through the PICDRP. The ceiling makes the commitment administrable and limited rather than an open-ended power to reserve an undefined universe of labels.

The fourth element is standing. ACTO and its member states are the sole entities entitled under Section 4 to invoke or contest an alleged failure by the operator to comply with the ACTO Commitment through the PICDRP. Standing is a form of power because it determines who can convert a suspected breach into a recognised dispute. It is also a boundary: the right concerns compliance with the specified commitment, not general supervision of the registry or a power to reopen the original delegation decision.

ICANN separately retains contractual enforcement authority. The agreement makes the ACTO Commitment enforceable by ICANN and through the PICDRP, and requires the operator to implement remedies imposed after a panel determination. The text expressly states that a reasonable remedy may include termination under the agreement. Termination is not automatic upon an allegation. The enforcement chain requires a cognisable obligation, a complaint or compliance action, a determination and the applicable opportunities to cure or challenge before the authorised contractual consequence can follow.

Specification 13 does not undo this architecture. It recognises the strings as dot-brand TLDs and supports an operator-centred registration model. The BAMC and Board treatment of Reconsideration Request 19-1 supports the coexistence of brand status and a separate Specification 11 safeguard package. Amazon obtained a brand-registry framework subject to the ACTO Commitment, not freedom from it.

The final terms also identify what ACTO did not obtain. The agreements do not make ACTO a co-registry operator. They do not require general ACTO consent for every registration, amendment, assignment or root-zone request. They do not give ACTO authority over DNSSEC, the EPP interface, registrar accreditation, nameserver administration or IANA processing. Its durable powers are the designated-name mechanics, the protected-term mechanisms and standing to use the specified dispute route.

That bounded design was the institutional compromise. Amazon retained operational unity for a dot-brand registry. ACTO obtained protections capable of surviving the negotiating table because they were attached to three agreements. The Board avoided treating either governmental concern or the applicant’s brand claim as conclusive by itself. ICANN organisation received terms it could administer through an existing contractual framework.

A contract proves what it contains and what the enforcement system can reach. It does not, by itself, prove that ACTO later submitted the full protected-term list, that all nine names in each TLD were requested or activated, or that a complaint was filed. A targeted search of published ICANN records through 31 July 2026 did not locate a public PICDRP determination, formal cure notice or published contractual enforcement action under the ACTO Commitment. That absence is not proof that no request, informal compliance exchange or non-public interaction occurred.

The supported conclusion is that an enforceable mechanism existed, not that every mechanism was used or tested.

Signature, assignment and delegation were separate gates

The executed agreements did not put the strings into the root on 19 December 2019. Section 1.1 of the .AMAZON agreement, mirrored in the IDN agreements, designated the registry operator subject to the requirements and necessary approvals for delegation and entry into the root zone. ICANN’s same-day status notice likewise stated that pre-delegation requirements remained.

That reservation preserves a division of institutional labour. Contracting determines who is bound to operate under ICANN’s registry rules and what obligations attach to that role. Pre-Delegation Testing and String Delegation Readiness determine whether the contracted application and technical arrangements satisfy programme prerequisites. Root-zone management then verifies an eligible request from the correct party and implements delegation. A valid agreement can fail to reach the root if later conditions are not met.

Before the final delegation stage, the agreements were assigned. ICANN’s completed assignment register records 14 February 2020 as the date on which all three moved from Amazon EU S.à r.l. to Amazon Registry Services, Inc. A published assignment instrument for the Chinese variant shows the legal technique: the assignor transferred all rights and obligations under the Registry Agreement, and the assignee accepted the assignment, assumed all related liabilities, whether contingent or accrued, and agreed to perform the assignor’s covenants and obligations. The register establishes the same transfer date and parties for all three agreements. The published evidence does not support adding variant-specific terms beyond the instrument available for xn--jlq480n2rg.

The assignment did not strip out the ACTO Commitment. The duties were part of the agreements whose rights, obligations and liabilities moved to the assignee. Nor did assignment itself establish readiness. It identified the recognised contractual party that would have to satisfy the remaining programme and delegation checks.

The public readiness record is now specific. The 28 May 2020 .AMAZON String Delegation Readiness report names Amazon Registry Services, Inc., states “Readiness Confirmed” and records successful Pre-Delegation Testing. Parallel reports do the same for application 1-1318-83995 and application 1-1318-5591. Those reports eliminate the need to infer successful testing solely from the later IANA delegation documents.

The current IANA root records give 28 May 2020 as the registration date for .AMAZON, xn--cckwcxetd and xn--jlq480n2rg. IANA’s separate reports published on 2 June document the delegation review for .AMAZON, the Japanese variant and the Chinese variant. They record eligibility, consistency between the approved or contracted party and proposed manager, administrative and technical contacts, technical conformance and completion of required processing.

The date distinction is useful. The readiness reports and root registration date are 28 May; the public IANA review reports are dated 2 June. It would be imprecise to treat 2 June as the only possible date attached to operational delegation, just as it would be imprecise to treat the 19 December contract date as root insertion. The records describe related but different administrative events.

Those technical checks made delegation executable. They did not create another forum for deciding whether the strings had cultural significance or whether the Board had correctly weighed GAC advice. Reopening those questions at implementation would have transformed root-zone management into an unbounded merits appeal. Its role was to confirm that the policy and contractual chain had produced an eligible, technically conformant request from the recognised party.

Current records checked through 31 July 2026 preserve the resulting operator chain. ICANN’s agreement profiles for .AMAZON, xn--cckwcxetd and xn--jlq480n2rg each identify Amazon Registry Services, Inc. as operator under a 19 December 2019 Base, Brand (Specification 13), Non-Sponsored agreement. The corresponding IANA records list the same company as sponsoring organisation.

The amendment record also matters. The 2023 Global Amendment applied to all three agreements and changed generic Base Agreement provisions and the introductory enforcement paragraphs of Specification 11. The 2024 Global Amendment also applied to all three, modifying DNS-abuse provisions in Specification 6 and Section 3(b) of Specification 11. Neither published amendment rewrote the bespoke Section 4 ACTO Commitment. The current record therefore supports continuity of that commitment while leaving its later exercise and enforcement as separate factual questions.

What the public record supports—and what it does not

The principal chronology is unusually well documented. Application records, the full Community Objection determination, NGPC resolutions, reconsideration records, the divided IRP declaration, later Board resolutions, the applicant’s proposal, public-comment and contracting notices, executed agreements, assignment records, readiness reports and IANA documents identify the main decision-makers and the acts they performed. The evidence supports a high-confidence conclusion about the architecture of authority.

Two points are supported by complete primary records. The full Independent Objector determination permits a precise account of the four cumulative Community Objection tests and why the objections failed. The public String Delegation Readiness reports provide exact report dates, identify the assignee as applicant at that stage, confirm readiness and record successful Pre-Delegation Testing. Both records make the institutional boundaries more concrete.

Other propositions remain unproven. The agreements establish rights and duties concerning designated names, protected terms, rejection, challenge and dispute standing, but the cited public record does not establish whether ACTO submitted a full culturally significant term list within the contractual window, which permitted names were requested in each TLD, whether any request was rejected or whether ACTO challenged a rejection. Those operational events cannot be inferred from the contract’s existence.

The same caution applies to enforcement. No published official case located in the targeted search records a PICDRP determination, formal cure notice or public remediation action under the ACTO Commitment. That does not establish perfect compliance, abandonment or non-use. A mechanism can exist without producing a public reported case, and informal or confidential interactions may not appear in the sources examined.

The current profiles and global amendments support continuity of Amazon Registry Services, Inc. as operator and show no published amendment to the bespoke Section 4 commitment through 31 July 2026. They do not prove how every obligation has been performed in practice.

The bounded conclusion is therefore stronger than a generic statement of compromise but narrower than a claim of complete implementation: the 2019 agreements created enforceable safeguards; the February 2020 assignments placed the agreements with Amazon Registry Services, Inc.; readiness and root-zone processing produced delegation; and the later public record does not show that the ACTO Commitment has been removed. Exercise and enforcement of each safeguard remain fact-specific questions.

A chain of necessary acts, not a single authorisation

The .AMAZON result can be tested through four counterfactuals.

Without the 2017 IRP declaration, the 2014 “Will Not Proceed” decisions could have remained in force. Yet without a later Board decision, the declaration itself would have left the applications where they were. Review changed the legitimacy and durability of the earlier decision, not the root zone.

Without the Board’s re-evaluation and its October 2018 removal of “Will Not Proceed” status, the Abu Dhabi negotiating route could not have become resumed application processing. Yet resumed processing still left the Board free to reject an inadequate proposal and left ICANN organisation unable to contract until the remaining programme work was complete.

Without publication, comment and incorporation into Specification 11, Amazon’s safeguards would have remained political assurances. Yet the executed agreements still did not delegate the strings. Assignment, successful Pre-Delegation Testing, confirmed readiness and root-zone processing remained necessary.

Without the delegation pathway, a valid agreement with a technically unready or mismatched operator would not have produced a root entry. Yet IANA did not supply the public-policy legitimacy of the arrangement. Its review relied on the completed policy, contracting and operator-identification chain.

No institution therefore “authorised .AMAZON” in the comprehensive sense implied by the question. Governments created public-policy pressure and privileged advice. The GAC altered the Board’s procedural burden. The NGPC made the 2014 stop decision. The IRP majority declared the decision-making inconsistent with ICANN’s obligations and recommended re-evaluation. The Board reopened the applications, supervised the negotiating timetable and judged the safeguard proposal sufficient. ACTO shaped specific protections. ICANN organisation administered change, comment and contracting. The Registry Agreements made selected protections enforceable.

Assignment identified the eventual obligated operator. Readiness and IANA root-zone management completed the operational pathway.

The Community Objection supplies a parallel caution. Amazon’s success there was real, but it resolved only whether the Independent Objector had met the Guidebook’s cumulative test. It did not neutralise GAC advice. The IRP victory was also real, but it corrected how ICANN decided rather than awarding the strings. The ACTO Commitment was enforceable, but did not make ACTO the registry operator. Each outcome mattered within the institutional entity it governed and no further.

That segmented structure is the case’s central governance fact. It prevented one actor from controlling application, objection, advice, review, bargaining, contract and root implementation. It also created opportunities for responsibility to blur: the Board could hide behind advice, governments could describe influence as consent, an applicant could describe review success as entitlement, a negotiating phrase about shared governance could be mistaken for joint technical control, and a contract could be mistaken for a live root entry.

The record becomes intelligible only when each act is attached to the institution capable of making it and to the remedy that act could actually produce.