AssemblyAI incident
Increased error rates for pre-recorded Speech-to-Text
AssemblyAI experienced a minor incident on September 30, 2025 affecting Asynchronous API, lasting 2h 54m. The incident has been resolved; the full update timeline is below.
Affected components
Update timeline
- investigating Sep 30, 2025, 05:33 PM UTC
Our team is currently investigating an issue that is causing an increase in server errors for our pre-recorded audio speech-to-text feature. This issue is only affecting a small number of requests, but it is affecting requests to both Universal and Slam-1 models. If you experience a server error we suggest retrying the request as those will likely be successful.
- identified Sep 30, 2025, 06:25 PM UTC
Our team has identified where the errors are occurring in our pipeline and is currently working to identify and resolve the root cause. This issue is currently only affecting an extremely small percentage of total pre-recorded audio traffic. 99%+ of requests are still completing successfully.
- monitoring Sep 30, 2025, 07:28 PM UTC
We have released a fix to address the small increase in errors we were seeing and are monitoring to ensure continued good performance.
- resolved Sep 30, 2025, 08:28 PM UTC
We have continued to see good performance for pre-recorded audio so we are now marking this issue as resolved.