Security Incident Response Policy
Version 1.0 of 3 November 2026
This English version is provided for convenience. The French version is the only legally binding version.
1. Purpose and scope
This policy describes how Unyel, a service published by Ethan Jordan, sole proprietor (entrepreneur individuel) trading as Ethaxis, based in Osny (France), detects, handles and documents security incidents. It covers Unyel's systems (website, API, MCP server, database, files, server functions), the administration accounts held with providers and all data processed, whether Unyel acts as controller or as processor for its customers.
A security incident is any event that compromises or may compromise the confidentiality, integrity or availability of these systems or this data. A personal data breach is an incident leading to the accidental or unlawful destruction, loss, alteration, unauthorised disclosure of, or access to, personal data (Article 4(12) GDPR).
2. Roles
The publisher is responsible for incident response: detection, assessment, containment, notifications and follow-up. Contact: support@unyel.com.
3. Detection
The detection sources in place are:
- automated monitoring of Unyel every 30 minutes, read-only on the database: rate counters per key and per address, authentication failures, monthly quotas, workspace creation, failed incoming events, audit log volume; any alert is sent immediately to the publisher's phone;
- daily database backup, which sends an alert if it fails;
- daily morning health check: status of Supabase projects, latest Vercel deployments and domain responses, scheduled tasks;
- audit log of sensitive actions (account connections and disconnections, keys, authorisations, workspace deletions, writes);
- Supabase server function logs and security advisors, Vercel runtime logs;
- security email sent to the customer for every new access granted to its workspace, allowing it to report an access it did not grant;
- reports received at support@unyel.com from customers, researchers or third parties;
- notifications from platforms (Shopify, Meta, Google and others) and providers (Supabase, Vercel, Brevo, OVH).
4. Assessment and severity
Each signal is examined upon receipt. The publisher determines whether it is an incident, which systems, data, customers and individuals are affected, and assigns a severity level:
| Level | Definition | Examples | Response time |
|---|---|---|---|
| Critical | Confirmed or likely unauthorised access to personal data or secrets | Leak of a service key or of the token encryption key, one workspace reading another's data, compromised administration account | Immediate |
| High | Exploitable flaw with no evidence of exploitation, or general outage | Isolation flaw discovered, platform token exposed, service offline | Within 4 hours |
| Medium | Flaw hard to exploit, abuse limited to one workspace | Quota abuse, blocked brute-force attempt, misconfiguration without exposure | Within 24 hours |
| Low | Event with no effect on data | Harmless traffic spike, unfounded alert | Within a week |
When in doubt, the highest level applies. The level is reviewed as the facts are established.
The time at which the publisher is reasonably certain that a data breach has occurred is recorded: it starts the notification deadlines of section 8.
5. Containment
Depending on the incident, the publisher takes one or more of the following measures, starting with those that stop the exposure:
- revoke the API keys and AI application access grants concerned;
- disconnect the accounts and stores concerned, which erases their tokens; ask customers to revoke access on the platform side (for Shopify, uninstall the app or change its secret);
- rotate server secrets (Supabase service keys, token encryption key, platform app secrets, scheduler key); rotating the encryption key makes all stored tokens unreadable, which requires accounts to be reconnected;
- replace a server function with a version that refuses requests, or pause the project;
- roll the website back to an earlier Vercel deployment;
- change passwords and close sessions on administration accounts (Supabase, Vercel, GitHub, Shopify Partners, OVH, Brevo);
- preserve evidence before any fix: copies of the relevant audit log rows, function logs and deployments.
6. Eradication
The publisher identifies the cause (code flaw, configuration, exposed secret, compromised account), fixes it, adds an automated test that replays the attack and checks that no other part of the service has the same defect.
7. Recovery
The service is restored once the cause has been fixed and verified. If data has been altered or lost, it is restored from the most recent daily backup taken before the incident. The publisher monitors the service closely in the following days and confirms the return to normal to the customers concerned.
8. Notifications
Customers (merchants): when a breach affects data that Unyel processes on behalf of a customer, the customer is notified by email without undue delay and no later than 48 hours after awareness, with the information set out in the data processing agreement, so that it can meet its own 72-hour deadline. The customer alone decides whether to notify the supervisory authority and its own customers for that data.
Shopify: any actual or suspected breach of a Shopify merchant's data is reported to Shopify Partner Support (https://help.shopify.com/questions/partners) immediately and no later than 24 hours after awareness, as required by section 9.14 of the Shopify Partner Program Agreement.
Other platforms: Meta, Google, TikTok and the other platforms concerned are notified under their own developer terms.
CNIL: where Unyel is the controller (Unyel account data, logs), any breach likely to result in a risk to individuals' rights and freedoms is notified to the CNIL without undue delay and, where feasible, no later than 72 hours after awareness (Article 33 GDPR), through the CNIL's online notification service. If not all information is available, the notification is made in phases. Any delay beyond 72 hours is justified.
Data subjects: when the breach is likely to result in a high risk to their rights and freedoms, data subjects are informed without undue delay, in clear and plain language, of the nature of the breach, its likely consequences, the measures taken and the contact point (Article 34 GDPR), except where Article 34(3) exempts this.
9. Incident log
Every incident, including those not notified, is recorded in the incident log kept in the project repository, in the file docs/journal-incidents.md: facts, data and individuals concerned, severity, effects, measures taken, notifications made or reason for not notifying (Article 33(5) GDPR). The log is kept for at least five years.
10. Post-incident review
Within 15 days after closing an incident of medium severity or higher, the publisher writes a review: timeline, cause, what worked, what must change. The decided corrections are carried over into the security documentation and, where necessary, into this policy.
11. Annual test
This policy is tested at least once a year through an exercise based on a fictitious scenario (for example the leak of a service key), which checks the detection, containment and notification steps. The exercise and its conclusions are recorded in the incident log, marked "exercise". The policy is reviewed on that occasion.
12. Contact
To report an incident or a vulnerability: support@unyel.com.