23 days ago
The subprocessor list at trust.railway.com includes entries we would like to understand in relation to Customer Data under the DPA:
PlanetScale – "Managed Postgres database hosting – US" (added October 5, 2026): Which service does this relate to? Are standard Railway Postgres databases (including ours in Amsterdam), their backups, or point-in-time recovery data hosted on or replicated to PlanetScale? If so, in which region?
OpenAI, Anthropic, SpaceXAI (via Anysphere) and Anysphere (Cursor) – "Generative AI services": For which purposes does Railway use them (e.g. Railway Agent, support, abuse detection, internal tooling)? Do they receive Customer Data such as logs, metrics, environment variables, source code or database contents, and only when a customer actively uses the Railway Agent? Is the data used for model training, and how long is it retained? Is there an opt-out at workspace or project level?
Regarding the security disclosure of October 21, 2026: Is our workspace among the affected accounts?3 Replies
Status changed to Awaiting Railway Response Railway • 23 days ago
23 days ago
On your Postgres data: native volume backups and PostgreSQL point-in-time recovery data are stored in the region your service deploys to, so a database deployed to EU West (Amsterdam) has its backups and PITR archives in EU West as well. Logs are the exception, as all logs are stored in US West regardless of deploy region. The published list at trust.railway.com and our DPA remain the authoritative record of our subprocessors.
On the October 21 incident, we're sorry, but we're unable to determine whether an individual workspace's variables were accessed or its credentials used. We understand that's frustrating when you're asking about your own workspace, and we're sorry for the worry and extra work this has caused. Out of an abundance of caution, please rotate credentials stored in Railway before the incident, update the variables, and redeploy the services that use them. Reviewing your providers' activity logs can also help identify recorded misuse. Our preliminary incident report is the written account of the incident you can use in your review.
Status changed to Awaiting User Response Railway • 23 days ago
brody
On your Postgres data: native volume backups and PostgreSQL point-in-time recovery data are stored in the region your service deploys to, so a database deployed to EU West (Amsterdam) has its backups and PITR archives in EU West as well. Logs are the exception, as all logs are stored in US West regardless of deploy region. The published list at [trust.railway.com](https://trust.railway.com/item/subprocessors) and our [DPA](https://railway.com/legal/dpa) remain the authoritative record of our subprocessors. On the October 21 incident, we're sorry, but we're unable to determine whether an individual workspace's variables were accessed or its credentials used. We understand that's frustrating when you're asking about your own workspace, and we're sorry for the worry and extra work this has caused. Out of an abundance of caution, please rotate credentials stored in Railway before the incident, update the variables, and redeploy the services that use them. Reviewing your providers' activity logs can also help identify recorded misuse. Our [preliminary incident report](https://blog.railway.com/p/preliminary-report-october-security-incident) is the written account of the incident you can use in your review.
23 days ago
Thank you, that is helpful. Could you please also confirm:
-
PlanetScale: Is any Customer Data of a standard Railway Postgres service
deployed to EU West (the database itself, replicas, backups or PITR archives)
hosted on or replicated to PlanetScale? A short yes/no is enough.
-
Generative AI subprocessors (OpenAI, Anthropic, SpaceXAI via Anysphere,
Anysphere/Cursor): For which purposes are they used, and do they receive
Customer Data (logs, metrics, environment variables, source code, database
contents) of customers who do not actively use the Railway Agent? If they do:
is it used for training, how long is it retained, and is there an opt-out?
-
Logs in US West: Does this cover both the HTTP/edge logs and the deployment
(stdout/stderr) logs, and how long are they retained?
Status changed to Awaiting Railway Response Railway • 23 days ago
23 days ago
No. PlanetScale is used for our internal systems and is not customer database hosting. A standard Postgres service deployed to EU West, including its database contents, replicas, volume backups and PITR archives, is not hosted on or replicated to PlanetScale.
We use the generative AI services for build debugging and support threads, so their use is not limited to customers who actively use the Railway Agent. Relevant build or support context may be processed by a model provider as part of those workflows, though not every listed provider takes part in every workflow. Customer data is not used for model training in any capacity, by us or by our model providers. We can't give you a provider-side retention period or a workspace or project opt-out here. The subprocessor list and the DPA remain the published record.
Yes, US West log storage covers both HTTP/edge logs and deployment stdout/stderr logs, along with build, deploy, DNS and network logs. All of them are retained for 90 days and then deleted. The shorter figure for your plan (7 days on Hobby) only limits how far back the dashboard shows logs. It is not the retention period.
Status changed to Awaiting User Response Railway • 23 days ago
16 days ago
This thread has been marked as solved automatically due to a lack of recent activity. Please re-open this thread or create a new one if you require further assistance. Thank you!
Status changed to Solved Railway • 16 days ago