超過一個月客訴 【Cross-Device Chat Data Missing From Official Exports - Support Unresponsive】

ChatGPT Bugs
1

I am a long-term paying user of ChatGPT and Codex. Since July, I have been reporting a serious issue involving inconsistent states of the same long-running conversation across the Android App, web interface, shared page, and official OpenAI data exports.

I have already performed a substantial amount of technical investigation that should not normally be the responsibility of an end user, including:

  • Comparing the Android App and web versions

  • Requesting and verifying multiple official data exports

  • Examining the conversation ID, current node, and message nodes

  • Manually preserving records still visible on Android

  • Performing SHA-256 integrity verification

  • Preparing a complete case timeline, screenshots, and evidence package

  • Using additional paid Codex capacity to investigate an OpenAI product issue

The verified facts currently include:

  1. The Android App continued to display newer conversation content through at least July 19, 2026. I independently preserved and verified that material.

  2. The actual message content in OpenAI’s official exports ends on May 22, 2026.

  3. The exported metadata nevertheless indicates that the conversation was updated through July 11, while the exported message graph contains no messages after May 22.

  4. Official exports generated at different times contain the same omission. This does not appear to be adequately explained as a one-time export failure.

  5. A share result generated from the original Android conversation displays another title, “薑汁汽水 - 不純*聊天室,” or resolves to a different conversation state. OpenAI has not explained how this title, the related conversations, branch, current node, and share snapshot were created or connected.

  6. My support history now involves multiple case numbers. I also received a “New Case” communication without a clear explanation of its origin or relationship to the existing case history.

  7. OpenAI Support stated that the matter would be prepared for “specialized technical review,” but has not confirmed:

    • Whether the technical review was actually opened

    • The date it was opened

    • Which team currently owns it

    • Whether all evidence and attachments were incorporated

    • Any internal tracking reference

    • The date of the next substantive update

  8. I will be changing phones next week. The old Android device is currently the only directly observable client displaying the newer conversation state, making preservation time-sensitive. Support has nevertheless remained silent.

  9. Today, while using ChatGPT normally for a work-related discussion, the service suddenly displayed:

    “Too many requests. Your requests are too frequent. We have temporarily restricted your access to conversations. Please try again in a few minutes.”

    The notice did not disclose the trigger, scope, applicable usage limit, expected duration, appeal route, or whether the restriction affects my account standing or future access.

I understand that online services may apply rate limits, and I understand that technical investigations take time.

What I do not accept is:

  • Restricting a paying customer without identifying the applicable rule or expected restoration time

  • Official exports omitting records still visible in the App without a technical explanation

  • The same conversation resolving to different states across clients and sharing surfaces

  • Fragmenting support history across cases without a clear audit trail

  • Requiring the customer to perform preservation, hashing, node inspection, and cross-client comparison

  • Consuming the customer’s time and paid Codex capacity without addressing compensation

I am requesting that OpenAI:

  1. Keep the primary case open and cross-reference every related case.

  2. Confirm whether the specialized technical review was formally opened, identify the responsible team, and provide the next update date.

  3. Preserve the relevant conversation, branches, nodes, attachments, metadata, recoverable backup records, and diagnostic logs to the extent available.

  4. Provide the Android-visible records missing from the official exports.

  5. Explain why Android, web, the shared page, and official exports resolve to different states.

  6. Explain the provenance of “a different and unexpected conversation title” and its technical relationship to the original conversation.

  7. Explain the basis, scope, duration, and account impact of the new paid-service access restriction.

  8. Provide reasonable compensation, including the affected ChatGPT subscription period, restoration of Codex usage consumed during the investigation, and appropriate service credits.

I am also preparing a formal Better Business Bureau complaint to establish an independent external record of the unresolved paid-service, data-integrity, support-process, and compensation issues.

I am not demanding that a support representative instantly solve a complex engineering problem. I am asking for basic case accountability: acknowledge the evidence, preserve the records, identify case ownership, provide a traceable technical status, and offer remediation for the service interruption and debugging burden transferred to the customer.

Has anyone else experienced any of the following?

  • The same conversation showing different content on Android and web

  • Official exports omitting messages still visible in the App

  • A shared link displaying a different title or conversation state

  • A sudden “Too many requests” restriction during normal paid use

  • A new support case being created without the previous history and attachments being clearly carried forward

If you have experienced something similar, please share the platform, date, App version, and whether Support ever provided a concrete technical explanation.

【付費 ChatGPT 使用者被迫自行替 OpenAI 做跨裝置資料 QA，客服仍未實質回覆】

我是一名長期付費使用 ChatGPT 與 Codex 的使用者。從 7 月開始，我一直在向 OpenAI 反映同一筆長期對話在 Android App、Web、分享頁及官方資料匯出之間出現不同狀態的問題。

到現在，我已經自行完成大量本來不應由一般使用者負責的技術核對工作，包括：

  • 比較 Android App 與 Web 顯示內容

  • 多次申請並核驗官方資料匯出

  • 檢查 conversation ID、current node 與訊息節點

  • 手動保存手機端仍可見的對話

  • 以 SHA-256 核驗保存資料

  • 整理完整案件時間線、截圖與證據附件

  • 額外使用付費 Codex 額度協助調查 OpenAI 自己的產品問題

目前可以確認的情況如下：

  1. Android App 當時仍可顯示至少截至 2026 年 7 月 19 日的較新對話內容，我也已自行保存並核驗。

  2. OpenAI 官方資料匯出的實際訊息內容卻只到 2026 年 5 月 22 日。

  3. 更矛盾的是，官方匯出中的中繼資料顯示這筆對話曾更新到 7 月 11 日，但訊息正文仍完全沒有 5 月 22 日以後的內容。

  4. 不同時間取得的官方匯出仍出現相同缺漏，並不是單次匯出失敗可以合理解釋的情況。

  5. 從 Android App 原始對話產生的分享結果，卻顯示另一個標題「薑汁汽水 - 不純*聊天室」或不同對話狀態。OpenAI 至今沒有說明這個標題、相關 conversation、branch、current node 與 share snapshot 究竟如何產生及互相關聯。

  6. 我的客服內容先後涉及多個 Case Number，甚至收到未清楚說明來源與關係的「New Case」通知。我已多次要求 OpenAI 說明主要案件、案件之間的關係、附件是否完整納入，以及技術升級是否真的建立，但至今仍沒有完整答案。

  7. 客服表示會準備進行「專門技術審查」，卻沒有確認：

    • 技術審查是否真的正式建立

    • 建立日期

    • 目前負責團隊

    • 內部追蹤編號

    • 下一次實質更新時間

  8. 我下週即將更換手機，而舊手機目前仍是唯一能直接觀察較新 Android 對話狀態的裝置。這代表資料保全具有明確時效性，但客服仍然整天沒有回覆。

  9. 今天我正常使用 ChatGPT 討論工作時，系統又突然顯示：
    「太多要求。你的要求過於頻繁。我們暫時限制了你的對話存取權限。」

    系統沒有告知觸發原因、限制範圍、適用上限、解除時間、申訴方式，也沒有說明是否影響帳號狀態與後續使用。

我理解線上服務可能有速率限制，也理解技術調查需要時間。

但我不能接受的是：

  • 付費使用者被限制服務，卻看不到適用規則與解除資訊

  • 官方匯出缺少 App 仍可見的資料，卻長期沒有技術解釋

  • 同一筆對話跨裝置及分享頁解析到不同狀態

  • 客服案件被拆分或另開，卻沒有清楚的案件治理與可追溯紀錄

  • 使用者被迫自行做資料保全、雜湊核驗與跨端比對

  • OpenAI 持續消耗使用者的時間與付費 Codex 額度，卻不回覆補償要求

我目前要求 OpenAI：

  1. 保持主要案件開啟，確認所有相關案件均已交叉關聯。

  2. 書面確認專門技術審查是否已正式建立、由哪個團隊負責，以及下一次更新日期。

  3. 保全相關 conversation、branch、node、附件、中繼資料、可復原備份及診斷紀錄。

  4. 提供 Android App 可見但官方匯出缺少的完整資料。

  5. 說明 Android、Web、分享頁及官方匯出為何顯示不同狀態。

  6. 說明「a different and unexpected conversation title」的來源及其與原始對話的技術關係。

  7. 說明這次付費服務限制的規則、範圍、期限及帳號影響。

  8. 提出合理補償，包括受影響期間的 ChatGPT 費用、調查過程消耗的 Codex 額度及服務點數。

我也準備向 Better Business Bureau 提交正式申訴，建立獨立的外部紀錄。

我不是要求客服立刻憑空修好所有問題。我要求的是最基本的案件責任：確認收到證據、維持資料完整、清楚交代案件歸屬、提供可追蹤的技術處理進度，並對付費服務中斷與使用者被迫承擔的除錯成本提出補救。

請問社團裡是否也有人遇過：

  • 手機 App 與 Web 顯示同一對話的不同內容

  • 官方匯出缺少 App 仍可見的訊息

  • 分享連結出現不同標題或對話狀態

  • 無預警出現「太多要求」並被暫時限制對話權限

  • 客服另開案件後，原案件歷程或附件沒有被完整承接

若有類似經驗，希望大家能提供發生平台、日期、App 版本，以及客服是否曾給出具體技術說明。