On this page
Zeabur has confirmed a second security breach on 14 and 15 October 2026 and is shutting down on 31 December. If anything of yours lives on Zeabur, replace every credential stored there, revoking each one at the provider that issued it, and then migrate before your service stops. The first action is the rotation, not the move. Both sets of dates come from Zeabur's own report and its shutdown notice.
| Fact | Detail | Source |
|---|---|---|
| What happened | An attacker used leaked internal credentials on 14 and 15 October 2026 (UTC) to reach the shared cluster, some Dedicated Servers and the platform database | Zeabur report |
| Report status | Preliminary, published 19 October 2026, with no revision and no newer Zeabur post on the incident at our 2 November check | Zeabur report |
| Platform database | 3,218,965 environment variable records and 321,810 accounts, which Zeabur marks as confirmed (a log records it) | Zeabur report |
| Tokens and keys | 566 zSend API keys stored in plaintext and 11,536 platform Access Tokens stored as hashes; all revoked | Zeabur report |
| Shared clusters | Commands were run in 865 database services. The logs show commands, not output, so what was taken is unknown | Zeabur report |
| GCP Dedicated Servers | 101 servers belonging to 92 users had attacker root sign-ins; Zeabur says to treat them as compromised | Zeabur report |
| Where it started | Initial credential source not confirmed. Zeabur names a leaked GitHub App key from August as the leading hypothesis, without direct evidence | Zeabur report |
| Use of the stolen data | What the database queries returned and how it was used later are unknown | Zeabur report |
| Shutdown | Announced 17 October. From 1 December renewals stop and shared clusters are paused. All services stop on 31 December | Shutdown notice |
| Independent verification | None found. Every figure here is Zeabur's own statement | this page |
| Checked | 2 November 2026 | this page |
Who is affected by the Zeabur breach?
Zeabur's report sorts customers by where their secrets were held. Read your row, then use the checklists below.
| If you used | What Zeabur's report says | Treat it as |
|---|---|---|
| Environment variables stored in Zeabur | 3,218,965 records in the platform database are marked confirmed | Exposed. Replace all |
| Shared cluster services (databases, n8n, new-api/one-api) | Secrets of services in 9 regions were read (partly confirmed), and commands ran in 865 database services | Exposed. Replace secrets and change database passwords inside the database |
| GCP Dedicated Server | Attacker root sign-ins on 101 servers | Compromised host. Rebuild |
| Hetzner or DigitalOcean Dedicated Server | Use of admin access cannot be confirmed | Check the provider console, then apply the GCP checklist |
| Variables moved to your own Dedicated Server after August | Zeabur says projects that moved their variables as advised in August avoided the platform database leak | Platform database not affected by that route. Still check the server if it was one of the listed hosts |
| A GitHub App connection | The app's private key was read on 15 October between 15:05 and 18:54 UTC; no use found so far | Check your GitHub audit log |
Zeabur also tells every user to act now and not to wait for an email (Zeabur report). The report says that of 890 Linode servers, the 847 with records show no attacker sign-ins. It says nothing about the other 43, so their status is unknown at our check on 2 November 2026. Ask support about them.
What is the trap in this one?
Two things catch people.
The first is treating migration as the fix. Zeabur's August notice said moving variables to a Dedicated Server "does not invalidate old credentials" (Zeabur, 7 September). The same holds for a move to any new host. A secret copied across unchanged is still the secret the attacker may have read.
The second is the shutdown clock. Zeabur's 22 October post recommends finishing migration by 30 November, says each dedicated server stops and its data is deleted on its own expiry date, and says all service data is deleted when services stop on 31 December (Zeabur, 22 October). The shutdown notice itself says 31 December does not mean every service runs until that day (Zeabur, 17 October). Look up your own server's expiry date today.
How do I fix it?
Zeabur's report lists the actions in a fixed order. This section follows that order and Zeabur's wording on scope.
What do I do if I stored variables or services on Zeabur?
- For every credential in a Zeabur environment variable, revoke the old one at the provider first, then create a new one and redeploy. Include API and cloud keys, database passwords and signing keys (Zeabur report).
- Sign in again, recreate your Zeabur Access Tokens, and look for requests you do not recognise.
- If you use zSend, recreate your zSend API keys.
- If you use the GitHub App, check the GitHub audit log for Zeabur App actions from 15 October 15:05 to 18:54 UTC: unexpected clones or pushes, new deploy keys, changed webhooks.
- Review bills from 14 October onward. GCP Dedicated Server users should look back to 1 September.
For shared-cluster services, Zeabur adds: change database passwords inside the databases, replace n8n encryption keys and recreate its third-party credentials, revoke upstream AI keys used by new-api/one-api, and check PostgreSQL for roles, functions, triggers, scheduled jobs or extensions you do not recognise. It says to treat backup data in the tpe1 and hkg1 regions as leaked.
What do I do if I ran a GCP, Hetzner or DigitalOcean Dedicated Server?
For GCP, Zeabur says to treat the server as compromised and to back up data and rebuild it, which it calls the most reliable fix. If you cannot rebuild at once, its minimum list is: remove unknown SSH keys from root, check sudo-capable accounts, cron entries, systemd services and k3s workloads, change the root password, disable SSH password login, replace every credential on the server, and check sign-in records back to 1 September (Zeabur report). For Hetzner and DigitalOcean, check in the provider console whether the root password was reset or the system reinstalled after 14 October, then apply the same list.
Send support this, and keep the reply:
For my account [email], which of my services and Dedicated Servers are in the set Zeabur lists as accessed on 14 and 15 October 2026? Please give the server expiry dates, and confirm what I should expect in the incident compensation process before the 30 November deadline.
What does the shutdown change, and what are the migration options?
Zeabur's dates, from the 17 October notice and the 22 October post:
| Date | What Zeabur says happens |
|---|---|
| Now | New dedicated server and domain purchases have stopped. Balance refund requests are open |
| 30 November 2026 | Recommended migration completion. Deadline for balance refund requests through support and for security incident claims |
| 1 December 2026 | Dedicated server renewals stop; shared-cluster services are paused and can be resumed by hand to back up data |
| Server expiry date (1 to 31 December) | That dedicated server stops and its data is deleted |
| 31 December 2026 | All services stop and service data is deleted. Support runs until this date |
The migration guide (20 October) names Railway, Coolify, k3s and Docker Compose as targets. We have not tested any of them for this page. It lists what needs a new owner: images, database backups (Zeabur's daily backups end with the shutdown), *.zeabur.app domains, injected variables, internal .zeabur.internal addresses, certificates, logs and team permissions. It also warns not to paste your Zeabur token into an AI agent chat, because the token can reach your whole account.
One caution about picking a target. Railway published its own disclosure on 21 October, saying an unauthorised party may have accessed encrypted environment variables and asking customers to rotate all variables (Railway, 21 October). Its newer preliminary report, dated 22 October and read at our check on 2 November 2026, says its forensic investigation is still ongoing, that at most 1.85% of workspaces had some environment variables potentially compromised (the separate figure of more than 100,000 counts users affected by upstream provider key resets), and that all customers who signed up before 21 October should rotate their keys stored on Railway. Its 26 October update calls the incident resolved for existing users, with new signups still paused. That is a reason to read its latest status before moving secrets there, not a verdict on it. If you already use Railway, its own rotation steps are on a separate page.
After the move, DNS records that still point at Zeabur become a loose end. Red's free scan checks for DNS records pointing at services you no longer own, among other things, and lists them from outside.
Am I affected? A three-state self-test
Am I affected? Where you stand right now
A scan from outside cannot see what Zeabur stored for you, so it cannot tell you whether your secrets were read. Zeabur's records and your own provider logs can.
- Detected You have a Zeabur account, services or Dedicated Servers, and Zeabur's report lists your kind of resource as accessed. Rotate everything stored there now, in Zeabur's order, and rebuild any GCP Dedicated Server.
- Unknown You are not sure what you stored on Zeabur, or whether a server is in the listed set. This is the common case. Zeabur treats data it cannot rule out as leaked. Inventory your variables, ask support the question above, and rotate.
- Confirmed by owner evidence You hold the provider-side revocation records for each old credential and a redeploy after the new ones went in. The rotation is done. The remaining question is whether the old keys were used, which your provider logs and bills answer.
No scan, ours included, moves you from "unknown" to "confirmed".
Was the stolen data used?
Zeabur's report does not say. It records what the database queries returned and how that was used as unknown, and it names one hypothesis it has not confirmed: an OpenAI key abused on 16 October between 01:50 and 02:56 UTC, possibly from leaked variables (Zeabur report). Look at your own provider dashboards for usage you did not make since 14 October. Zeabur says compensation for third-party API charges goes through a support ticket, with masked details and the related credentials revoked first.
Claims close on 30 November, and Zeabur asks for the affected accounts and services, the types of leaked credentials, and invoices or usage records from the incident period (Zeabur, 22 October).
What is known about how it happened?
Zeabur says long-lived admin credentials were stored in plaintext and could be used from the public internet, and that the cluster admin tokens had no expiry. It has restricted access to the admin interface, revoked tokens, rotated keys and removed the attacker's internal node. How the first credentials leaked is not confirmed (Zeabur report). The report is preliminary and may change. Open it before acting on any figure here.
Part of a wider run of October 2026 platform incidents is tracked on the October 2026 security tracker. For a role-by-role list of what to do this week, see the October 2026 security checklist. Another hosted-database incident with the same rotate-first logic is the TiDB Cloud breach, and the other incident write-ups cover the rest. If a breach reaches a client site, a plan for the first hours helps.
Questions people ask
Was my data stolen in the Zeabur breach?
It depends on where your secrets lived. Zeabur's report lists the environment variables of the platform database as read, shared-cluster services as accessed, and 101 GCP dedicated servers as signed into. For anything it could not rule out, Zeabur says it treats the data as leaked, and so should you.
I already rotated my keys after the August breach. Do I do it again?
Yes. Zeabur's report says to replace every password and key stored on Zeabur even if you replaced them after August. The October access was a separate incident, so the August rotation does not cover credentials added since or ones that were never rotated.
When does Zeabur shut down, and can I still get a refund?
Zeabur says dedicated servers stop renewing and shared clusters are paused from 1 December, and all services stop on 31 December 2026, with data deleted then. Its 22 October post sets 30 November as the deadline for balance refund requests through support and for security incident claims.
Will moving to another host fix the breach?
No. Moving gets you off a platform that is closing, but anything that was stored on Zeabur stays exposed until you revoke it at the provider that issued it. Rotate first or at the same time, and never copy old secrets into the new place unchanged.
Did the attacker shut Zeabur down?
Zeabur's founder announced the wind-down on 17 October and gives growth limits, system complexity and the two security incidents as the reasons. The company does not say the attacker closed it. The breach and the shutdown are two announcements you have to act on separately.
Sources
- Zeabur October 2026 security incident, preliminary report (Zeabur, 19 October 2026)The vendor's own account: timeline, counts, what is unknown, and the customer actions in Zeabur's order. Preliminary; Zeabur says it will update it.
- Zeabur is shutting down (Zeabur, 17 October 2026)The wind-down announcement and the 1 December and 31 December dates.
- Zeabur service migration, refunds, and security incident claims (Zeabur, 22 October 2026)Recommended migration date, data deletion, refund and incident-claim deadlines, support end date.
- Migrating off Zeabur: hand this guide to your coding agent (Zeabur, 20 October 2026)Names Railway, Coolify, k3s and Docker Compose as targets and lists what needs a new owner.
- A final check before leaving Zeabur (Zeabur, 23 October 2026)
- Zeabur August 2026 security incident (Zeabur, 7 September 2026)The first incident, and Zeabur's advice then on moving variables to Dedicated Servers.
- Railway security disclosure, 21 October 2026 (Railway)A separate incident at one of the migration targets Zeabur's guide names. Cited so you can weigh it.
- Preliminary Report: Oct 21st Security Incident (Railway, 22 October 2026)Railway's newer report: forensics ongoing, scope estimate and rotation advice.
Facts on this page are dated and point at the primary source next to them. If a vendor status changes after the date shown, the vendor's page wins; tell us at support@run.cloud and we will update it.