Some API requests may have experienced elevated error rates for approximately 2-3 minutes starting at 16:43 UTC on May 22, 2026. The issue has been resolved now.

Between 22:06 and 22:12 UTC, a subset of Voice Agent requests returned elevated 4xx errors following a recent deploy. The deploy has been rolled back and the service is operating normally.

This issue has been resolved.

This incident has been resolved.

This incident has been resolved.

This incident has been resolved.

This incident has been resolved.

A small number of customers experienced disruptions to STT fetch requests, callbacks, and third-party providers within Voice Agent, as well as 503 errors on API calls for usage lookups and token creation. The majority of customers were unaffected.

This incident has been resolved.

This incident has been resolved.

Between 15:16 and 15:32 UTC on March 25, 2026, we noticed an increased rate of API requests returning errors, including 403, 404, and 504 status codes. The issue has been identified and resolved.

This incident has been resolved

This incident has been resolved.

This incident has been resolved.

This incident has been resolved.

The upstream Anthropic incident affecting Claude models has been resolved and error rates have returned to normal. Voice Agent functionality should now be operating as expected.

A temporary issue during a Flux-related update caused intermittent SDK session errors. The issue was identified and resolved. 2:00 AM UTC - 13:45 UTC

As of 10:00 UTC, the issues from OpenAI models is being resolved.

This incident has been resolved.

The incident has been resolved.

This incident has been resolved.

The downstream provider indicated the issue was resolved. There was an outage between 16:20 UTC / 8:20 PT and 16:55 UTC / 8:55 PT

The incident has been resolved. The downstream provider's status page indicates the issues were resolved at 9:56 PT / 17:56 UTC

From 21:38 - 23:39 UTC on 2026-01-22, requests for Deepgram's Nova-3 multilingual STT model received degraded transcripts due to a faulty model release.

This incident has been resolved.

Informational (No Customer Action Required) Deepgram identified that our internal development environment was affected as part of the industry-wide Shai-Hulud 2.0 NPM supply chain attack. The attack used compromised NPM packages to inject malicious CI/CD workflows and attempt to exfiltrate internal development credentials. Our investigation confirms: - No access to customer data or databases - No impact to production API infrastructure or service availability - No modification of published Deepgram SDKs or packages - No effect on customer authentication or API keys Timeline (UTC): - 00:41, Nov 26: Webhook alert on an internal GitHub repo; engineers link activity to Shai-Hulud 2.0, disable malicious workflows, and rotate exposed credentials. - 21:45, Nov 26: Additional commits via a compromised GitHub App publish some internal materials (no customer or production impact); the account is removed, GitHub Actions disabled, and the org moved into a locked-down state. We see no further signs of compromise and are gradually restoring normal operations while tightening SDLC and CI/CD controls. Customer action: None required. Questions: security@deepgram.com • support@deepgram.com This incident affected internal development infrastructure only; customer-facing APIs and services were not impacted.

This incident has been resolved.

This incident has been resolved.

This issue has been resolved.

As of 19:28 UTC OpenAI's LLMs are no longer returning 500 errors. We are still in conversation with OpenAI about what happened.

This issue has been resolved.

As of 21:00 UTC on October 16th, OpenAI's LLMs are no longer returning 500 errors.

Start: 16:15:00 UTC End: 17:18:00 UTC Streaming connectivity issues from some AWS regions may have caused Websocket failures errors and degraded connectivity

This incident has been resolved.

Streaming and batch STT and TTS requests experienced elevated error rates due to connectivity issues in one of our regions that caused services to stop and not restart as expected. The issue has been resolved, and current status of the Deepgram API is operational.

Elevated error rates were observed for requests that submitted a URL to an input audio file or used callbacks to collect the response, primarily affecting STT batch requests.

This incident has been resolved.

This incident has been resolved.

This incident has been resolved.

Time: ~ 06:30 UTC - 07:00 UTC Summary: Streaming connectivity issues causing 'timed out during opening handshake' errors and degraded connectivity. Service is now stable. This has been resolved and root cause discovered.

Between 2025-08-14 21:13 UTC and 2025-08-14 21:37 UTC, the Voice Agent API experienced a 20-minute outage that prevented all new connections from being established. Existing active connections were not impacted. The issue has been resolved, and current status of the Voice Agent API is operational.

- Service degradation in inference infrastructure - High rate of premature streaming text-to-speech

- High rate of callback failures

- Service degradation in inference infrastructure - High rate of premature streaming text-to-speech

- Service degradation in inference infrastructure - High rate of premature streaming text-to-speech

This incident has been resolved.

This incident has been resolved. Error rates were generally within the window of 17:55 and 18:25 UTC.

This incident has been resolved.

A portion of streams connecting to Deepgram from 19:43 to 19:45 UTC resulted in 503 errors due to a deployment that was briefly in an unhealthy state.

Between March 22 at 20:00 UTC and March 24 at 14:39 UTC, some ASR requests that included redaction for non-English audio returned 500 errors due to an internal configuration issue. This issue affected only redaction requests and did not impact ASR performance for audio without redaction. Service has been fully restored for Deepgram models, and additional safeguards have been implemented to prevent similar incidents in the future.