Every incident, on the record.
This page is append-only. Anything that ever affected the confidentiality, integrity, or availability of KeyVault Edge gets documented here, honestly, with a post-mortem when useful.
Breach response playbook
If a breach affecting customer data occurs, this is the exact process we follow - published in advance so you can hold us to it.
T+0 - Detection
Automated anomaly detection or external report triggers an internal alert. The on-call engineer has 5 minutes to acknowledge. An incident channel is opened immediately.
T+15 min - Customer notification
Affected customers are emailed within 15 minutes of a confirmed breach affecting their tokens or keys. The email includes: what happened, what data was exposed, what we've done, and what you should do. We do not wait until we have all the answers - we send what we know and update as we learn more.
T+15 min - Automatic token revocation
All tokens involved in the incident are automatically revoked at the moment the breach is confirmed - no manual step required. Customers can re-issue clean tokens from the dashboard immediately.
T+1 hr - Containment
The affected code path, credential, or access vector is isolated. Cloudflare Worker deployments can be rolled back to a known-good hash in under two minutes.
T+24 hr - Forensic investigation
We conduct a full forensic review: what data was accessed, by whom, for how long, and what the entry point was. We commit to publishing a post-mortem on this page within 72 hours of containment.
T+72 hr - Public post-mortem
A full post-mortem is published here. It includes the timeline, root cause, affected scope, mitigation taken, and what we're changing to prevent recurrence. We use the format: what happened, why it happened, how we found it, what we did, what we're changing.
Credit policy
Any customer whose tokens were compromised in a confirmed breach receives a full billing credit for the affected period. We will not charge for a month in which we failed to protect your keys.
What we publish here
- Unauthorised access - attempted or successful - to any production system.
- Any incident where customer data may have been exposed, regardless of whether we can confirm it was.
- Downtime of the data plane (proxy) affecting more than 1% of customers for more than 5 minutes.
- Serious bugs found in the host-binding, token minting, or key rotation code paths.
- Sub-processor incidents that affect our service, with a link to the upstream report.
Log
No incidents to report.
No security incidents have been recorded since KeyVault Edge began operation. If something changes, it will be published here, not buried.
Think you've found something?
Responsible disclosure keeps everyone safer. Our policy covers scope, timelines, and public credit.
Read the disclosure policy →