BBB 投诉 OpenAI

把本页发给 Agent,并附上名、姓、邮箱、美国电话、五位邮编,写明「用这些信息创建账号并提交」。填写时使用用户本次提供的资料。

打开 https://www.bbb.org/file-a-complaint/1116-888255/form 一次,之后沿这份草稿的地址继续。国内访问使用和 OpenAI 相同的美国出口。用内置浏览器。

流程

  1. 填邮箱,点 Continue。验证码由用户提供,填入 #emailCode
  2. #firstName#lastName#phone#zip,点 Create Account
  3. 详情页把「过去 12 个月」选为 Yes,健康问题选 No。#problemDescription 贴下面的英文模板。订单号和金额留空。
  4. 诉求依次选择 Other (requires explanation)Billing adjustmentContact by the business。Explanation 填:Restore the affected account, correct routing so gpt-6-astra and gpt-6-sol are actually served, correct quota billing, and provide a written explanation.
  5. 点 Continue。
  6. 出现网址警告时勾选 I understand this warning and want to submit the form anyway.,保留 GitHub 链接,再点 Continue。
  7. 审核页勾选 agreeToTermsauthorizeCommunicationiConsentToTheCollection。核对姓名、邮箱、电话、邮编和正文后点 Submit。
  8. 成功页以 /thank-you 结尾,并显示 Your complaint has been submitted. 案件号约 48 小时内发到邮箱。出现图片验证码时把页面交给用户。

填写

英文模板

I am a ChatGPT Pro subscriber paying USD 200 per month. Recently the service became nearly unusable because the model intelligence dropped sharply. I used the open-source ModelTrace project (https://github.com/xqy2006/ModelTrace) to test the affected account. It uses independent numeric-sequence challenges. I compared the results with a friend's unaffected account, whose fingerprints were normal.

When I requested gpt-6-astra, repeated tests matched the fingerprint of gpt-5.6-luna instead, including several results above 99.9 percent and some above 99.99 percent. When I requested gpt-6-sol, repeated tests likewise matched gpt-5.6-luna rather than gpt-6-sol, again including results above 99.99 percent. One weaker gpt-6-sol run still ranked gpt-5.6-luna first and did not identify gpt-6-sol. My quota was still deducted as if gpt-6-astra or gpt-6-sol had been used.

I already submitted an appeal to OpenAI. The account has not been restored and the model routing has not been corrected. I am asking for restoration of the affected account, correction of model routing and quota billing, and a written explanation of what happened and how it will be prevented. I pay USD 200 per month and should not be charged for a higher model when a different model was actually served.