// Trust Center

Incident Response Framework

AuthorityRail's operational doctrine: Detection → Classification → Communication → Recovery. The same SEV ladder used internally is published here so customers and auditors can verify alignment with the response actions they observe.

Last revised: 2026-05-17 · Version v1.0

Severity ladder

SEV	Definition	Customer notification SLA	RCA SLA
SEV-1	Core execution authority surface broken or compromised. Mass DENY, signing failure, security incident with customer data exposure.	< 1 hour	5 business days
SEV-2	Single non-gate service down with degraded impact, or sustained gate performance degradation. Partial outage.	< 4 hours	10 business days
SEV-3	Single-customer impact, minor degradation, non-customer-facing internal issue.	< 24 hours (direct)	Internal log; published only if customer requests
SEV-4	Cosmetic, minor, no customer impact.	Monthly summary if relevant	None

How customers are notified

status.authorityrail.com — public status page; the first place to check for current state.
Targeted email from [email protected] for incidents affecting specific tenants.
Post-incident RCA published at trust.authorityrail.com/incident-reports/<YYYY-MM-DD>-<short-tag>.md within the SLA above.

Detection layer

Sentry application monitoring (per Sprint Closure #2).
Better Stack uptime monitoring on 8 customer-facing surfaces (per Sprint Closure #8).
PagerDuty on-call routing with continually-re-page-until-acknowledged for SEV-1 (per Sprint Closure #9).
Customer reports via [email protected] or [email protected].

Recovery layer — runbooks

Five critical runbooks are maintained against the highest-impact failure modes:

Signer failure (SEV-1)
Supabase outage (SEV-1 full / SEV-2 degraded)
Railway outage (SEV-1 / SEV-2)
Deployment rollback (SEV varies)
Customer key compromise (SEV-2 single / SEV-1 systemic)

Disaster recovery

Supabase Point-In-Time Recovery (PITR) targeted for activation per Sprint Closure #5 (Pro tier upgrade pending). Restore drill documented in docs/operations/disaster-recovery/supabase-pitr-restore-drill-2026-05-17.md.

Until PITR activation: Supabase managed backups on the AR-managed project. Multi-region active-active deployment is sequenced for Q1 2027 per the operational maturity gap analysis.

Postmortem doctrine

Every postmortem is blameless. The single on-call (today: Sammy Jones, founder) is one person; the goal is to make the system harder to break next time, not to assign fault. Postmortems use the canonical template at docs/operations/incident-response/rca-template.md.

Status page state mapping

SEV-1 → Major outage banner; affected component red.
SEV-2 → Partial outage banner; affected component amber.
SEV-3 (multi-customer) → Degraded performance; affected component amber.
SEV-4 → no status page state change.