AssemblyAI change history

AssemblyAI reported a major degradation of Universal-3.5 Pro async transcription on its US endpoint: slower processing and failed requests; EU was unaffected. The incident started at 2026-10-06 14:00 UTC and resolved at 2026-10-06 14:25 UTC (25 min), with impact reported beginning at 1:35p UTC. A fix restored normal processing; failed requests can be safely resubmitted.

AssemblyAI’s major EU asynchronous transcription degradation delayed Universal-3 Pro and Universal-3.5 Pro requests and caused some failures. It started at 2026-10-05 10:21 UTC and was resolved at 2026-10-05 11:39 UTC (1 h 18 min); processing returned to normal, and failed requests can be safely resubmitted. The US endpoint and real-time services were unaffected.

AssemblyAI recorded a critical US Async API outage starting 2026-09-16 20:22 UTC, lasting 1 h 14 min. At peak, roughly 50% of US async transcription requests failed; some successful requests were delayed, while EU and real-time/streaming services were unaffected. The vendor attributed the failure to AWS SQS and reported resolution after implementing a fix.

AssemblyAI reported a major Universal 3.5 Pro Realtime outage: EU endpoint users could not create streaming sessions and received code 3005 during 13:00–13:12 UTC; US users were unaffected. The incident started at 2026-08-19 13:17 UTC and resolved at 2026-08-19 13:24 UTC (7 min), after a fix restored session creation.

AssemblyAI reported a major US Asynchronous API outage starting 2026-06-25 15:37 UTC (reported impact began 15:00 UTC), resolved at 2026-06-25 16:48 UTC; computed duration: 1 h 10 min. Speaker diarization failures caused timeouts for speaker-labeled files under 2 minutes, with backlog delays and timeouts affecting other requests; a fix was released and processing recovered.

AssemblyAI reported a critical degradation affecting only EU pre-recorded audio transcription traffic for Universal-3-Pro and Universal-3-5-Pro. It started at 2026-06-24 09:13 UTC and was resolved at 2026-06-24 10:40 UTC, lasting 1 h 26 min. A fix was implemented at 10:05 UTC; all systems were operating normally at resolution.

AssemblyAI reported a major async transcription degradation from 2026-04-09 21:33 UTC to 2026-04-09 21:49 UTC, lasting 15 min. Some customers experienced increased processing times or throttled requests; AssemblyAI addressed a capacity constraint, cleared the transcription queue, and confirmed normal operation.

AssemblyAI reported a major billing incident: a small number of accounts received negative balances via the API, preventing transcription requests from completing. It ran from 2026-02-25 19:39 UTC to 2026-02-25 21:34 UTC (1 h 55 min). AssemblyAI attributed it to an unintentional error during its ongoing billing system migration, rolled out a fix, and confirmed resolution.

AssemblyAI’s Async Transcription and LLM Gateway APIs returned 5xx errors in a critical incident from 2026-01-29 17:54 UTC to 2026-01-29 19:32 UTC (1 h 38 min). Recovery involved elevated turnaround times while requests backlogged; AssemblyAI reported clearing the backlog and restoring normal turnaround times.

AssemblyAI reported a major degradation affecting Slam-1 in us-west-2: transcripts queued or errored; Universal was unaffected. The incident record started at 2025-11-20 08:21 UTC, though the vendor said delays began at 5:43 UTC. A fix restored normal processing at 08:47 UTC; resolution was confirmed at 09:41 UTC, with a recorded duration of 1 h 20 min.

AssemblyAI reported a major outage preventing documentation access and Google Authentication sign-in to the dashboard, attributed to an ongoing Cloudflare incident. It started at 2025-11-18 14:15 UTC and was resolved at 2025-11-18 20:58 UTC, lasting 6 h 43 min; access was reported restored at 15:15 UTC, with Cloudflare’s normal operation confirmed at resolution.

AssemblyAI reported a major partial outage affecting Slam-1 pre-recorded transcription requests, with slow processing, increased errors and failures. It started at 2025-10-22 17:56 UTC and was resolved at 2025-10-22 23:58 UTC, lasting 6 h 1 min. Universal requests were unaffected and offered as a workaround; Slam-1 performance returned to normal before resolution.

AssemblyAI reported a major outage: all transcription requests using Slam-1 failed; Universal was unaffected and offered as a workaround. The incident started at 2025-08-05 16:05 UTC and resolved at 2025-08-05 23:15 UTC (7 h 9 min), with successful requests observed again by 16:53 UTC. AssemblyAI made improvements to help Slam-1 scale more effectively.