Regarding the API key issuance API outage (August 21-23, 2026) [Resolved]
Thank you for your continued use of the AmiVoice API.
Between August 21 and August 23, 2026, some API key issuance API requests resulted in errors, preventing the issuance of API keys.
We sincerely apologize for the great inconvenience caused to our customers.
Period of occurrence
From approximately 18:04 on Friday, August 21, 2026 to approximately 19:38 on Sunday, August 23, 2026
Impact Range
In the API Key Issuance API (POST /issue_service_authorization), requests that did not specify the epi parameter resulted in an HTTP 500 error, causing a state where API keys could not be issued. The epi is an optional parameter that specifies the validity period of the API key to be issued, and when omitted, the default value of 30 seconds is applied.
- All requests that did not specify
epiwere failing during the period. - Requests specifying
epiare not affected.
No failures have occurred in the speech recognition API (Synchronous HTTP, WebSocket, Asynchronous HTTP), MyPage, or other services. However, customers using the API key issuance API were unable to obtain API keys, resulting in a state where speech recognition could not be initiated.
Cause
When system changes related to the API key issuance function were implemented, there was a defect in the processing when the epi parameter was omitted, resulting in a state where API key issuance failed.
In the preliminary verification, we were unable to sufficiently confirm the impact on request patterns that omit the epi parameter.
Recovery Response
The system change that caused the issue was reversed, and the system was restored around 19:38 on August 23, 2026.
After recovery, we have confirmed that the API key issuance API is operating normally, including cases where the epi parameter is omitted, and that similar errors have not occurred.
Measures to Prevent Recurrence
Regarding this incident, we are reviewing not only the individual implementation issues but also our methods for change management in the production environment, including pre-checking, post-change verification, and monitoring.
Details regarding measures to prevent recurrence will be reported separately after we have analyzed the causes and compiled the countermeasures.
We will strive to prevent similar incidents from occurring in the future and to ensure the stable operation of our services. We sincerely apologize again for the great inconvenience caused to our customers.