Reading
ElevenLabs' EU SIP Outage Is Fixed. The Cause Isn't Yet.
ElevenLabs' EU SIP calls failed for over two hours on Oct 6. It's marked resolved, but the bug behind it is still waiting on a permanent patch. Here's who was hit.
Prices last read 2026-10-07
Short answer
Not anymore. ElevenLabs' status page now marks the October 6 "EU Agents disruption in SIP calls" as resolved, and its last update says EU SIP services are "operational and stable."ElevenLabs published a write-up after the fix. It puts the impact at 2 hours 13 minutes, from 9:53 to 12:06 UTC on October 6, 2026. Affected traffic was SIP calls through EU SIP servers with static IPs. According to the write-up, about 66% of inbound calls and 100% of outbound calls on those servers crashed or timed out. The status page listed ElevenAgents and Telephony as affected, with the region shown as EU residency. When we checked on October 7, the main status page said ElevenLabs was not aware of any issues affecting its systems.Source: ElevenLabs Status – EU Agents disruption in SIP calls · 2026-10-07; ElevenLabs Status – Incident write-up · 2026-10-07; ElevenLabs Status · 2026-10-07
What actually broke: ports that never closed
ElevenLabs says a software bug during call teardown stopped network ports from being released back to the pool when calls ended. The servers ran for weeks without a restart, so the unreleased ports built up until they ran out. That caused connection timeouts and failed call setups. The fix has three parts: scheduled periodic restarts of the SIP servers to clear stale connections, new monitoring and alerts on port usage, and an audit of the call-termination code ahead of a permanent software patch.
Source: ElevenLabs Status – Incident write-up · 2026-10-07
Our take: credit where it's due. A resolved incident with a real root-cause write-up the same week is better than many voice-AI vendors manage. But look at the fix. Scheduled restarts are a workaround for a leak, not a cure, and the code patch is still described as coming. If you run production phone agents on the EU static-IP servers, the status page saying "resolved" shouldn't be the end of your interest. Watch for the permanent fix.
Region means your deployment, not your caller
The "EU" in this incident refers to ElevenLabs infrastructure, not to where callers are. The write-up limits the impact to EU SIP servers with static IPs. In ElevenLabs' SIP trunking docs, the data residency SIP endpoints follow the pattern sip-static.rtc.<region>.residency.elevenlabs.io, with eu as one of the region codes. The standard endpoint is sip.rtc.elevenlabs.io. Data residency is an Enterprise feature, and the EU environment is a separate account with its own workspace and API keys, at eu.residency.elevenlabs.io.
Source: ElevenLabs Status – Incident write-up · 2026-10-07; ElevenLabs Docs – SIP trunking · 2026-10-07; ElevenLabs Docs – Data residency · 2026-10-07
Quick self-check
Look at the hostname your SIP trunk points to. If it's the EU residency static endpoint, you were in the affected group. A European caller dialling an agent on the standard endpoint wasn't automatically in scope. A US caller reaching an agent on the EU residency environment could have been. Don't move to a different residency because of one incident. Residency is a data-location and compliance choice, and the environment is a separate account.
Still seeing failures? Work through the config checklist
The status page reports the incident as over. If your calls are still failing, ElevenLabs' SIP trunking docs are the place to troubleshoot. Treat this as a separate checklist, not as a sign the incident is still going.
Source: ElevenLabs Status – EU Agents disruption in SIP calls · 2026-10-07
Official SIP checks from ElevenLabs' docs
- Check the SIP trunk settings on both sides, in ElevenLabs and at your provider.
- Make sure your firewall allows SIP signalling on your ports: TCP 5060, or TLS 5061 (recommended for production). UDP 5060 is experimental and for testing only.
- For TLS, confirm your provider supports TLS on port 5061, uses TLS 1.2 or higher, and has a valid, trusted certificate chain.
- With digest authentication, recheck the trunk username and password and read your provider's logs for the exact error.
- For audio problems, allow UDP RTP traffic (typically ports 10000–60000) and confirm codec support: G.722 or G.711 PCMU/PCMA.
Source: ElevenLabs Docs – SIP trunking · 2026-10-07
are provided 'as is' and 'as available' without warranties of any kind
ElevenLabs Terms of Use, Section 10 (Disclaimers), last updated 31 March 2026Source: ElevenLabs Terms of Use · 2026-10-07
Why this clause matters: under ElevenLabs' general Terms of Use, the services are provided without a guarantee of uninterrupted availability. Any uptime commitment you rely on would have to come from your own agreement with ElevenLabs, not from the standard terms quoted here.
Source: ElevenLabs Terms of Use · 2026-10-07
The bigger lesson is about phone agents generally. With a chatbot, a 2-hour outage means a spinning loader. With a phone agent, it means calls that ring out or drop. If voice agents handle real customer calls for you, have a fallback route at your carrier, such as overflow to a human queue or voicemail, that doesn't depend on any single AI vendor staying up.
Optional: see a seller listing for ElevenLabs plans (third-party listing, not official access or an endorsement). That link carries our partner code Sk52g, which is how a purchase is attributed to this site. Promotion code UPD3S.
Sources (all read 2026-10-07)