Service Level Agreement

Service Tiers

EVE AI Core offers three commercially licensed tiers, each with distinct availability, performance, and support commitments. The SLA applicable to your deployment is determined by your executed license agreement.

Uptime Definition

“Monthly Uptime Percentage” is calculated as the total number of minutes in a calendar month minus the number of minutes of Downtime, divided by the total number of minutes in that calendar month, expressed as a percentage.

2.1 — What Constitutes Downtime

Downtime is defined as a period during which the governance decision pipeline is unable to accept and return verdicts for inbound governance requests. Specifically:

• The Authority Resolution API returns HTTP 5xx errors for more than three (3) consecutive minutes;
• The authority pipeline fails to return a verdict (PASS, VETO, or ESCALATE) within 30 seconds for any properly formed request; or
• The Cryptographic Authority Chain Layer is unable to persist verdict records, rendering the pipeline non-compliant with audit requirements.

2.2 — What Does Not Constitute Downtime

• Degraded performance of a single module (e.g., one CRD scorer returning cached results) while the overall pipeline continues to return verdicts;
• Scheduled maintenance windows communicated at least seventy-two (72) hours in advance;
• Unavailability of non-governance features, including but not limited to dashboards, reporting UIs, or webhook delivery;
• Latency increases that remain within the contractual p95 target for the applicable tier; or
• Issues caused by Customer systems, network connectivity, or third-party dependencies outside EVE AI Core infrastructure.

Latency Guarantees

Authority resolution pipeline latency is measured from request ingress at the Authority Resolution API to verdict egress. All measurements are collected via server-side instrumentation at one-second granularity and aggregated over rolling five-minute windows.

Latency targets apply to the governance decision pipeline only. They do not include:

• Network transit time between Customer infrastructure and EVE AI Core endpoints;
• Upstream LLM inference latency;
• Customer-side request serialization or deserialization; or
• Optional post-verdict webhook delivery.

Incident Classification

All incidents are classified by severity upon detection. Classification determines response time commitments, escalation procedures, and communication cadence.

Response Times

“Response Time” is defined as the elapsed time between EVE AI Core becoming aware of an incident (via automated monitoring or Customer report) and the first substantive communication to affected Customers. Automated acknowledgment emails do not satisfy this requirement.

5.1 — Communication During Active Incidents

• P1 incidents: Status updates every thirty (30) minutes until resolution or downgrade.
• P2 incidents: Status updates every two (2) hours during business hours.
• P3 incidents: Daily status updates until resolution.
• P4 incidents: Resolution notification upon completion.

5.2 — Post-Incident Reports

For all P1 and P2 incidents, EVE AI Core will deliver a Root Cause Analysis (RCA) report within five (5) business days of incident resolution. The RCA will include timeline of events, root cause identification, customer impact assessment, and corrective actions with implementation timeline.

Escalation Path

EVE AI Core operates a tiered escalation framework to ensure incidents receive attention commensurate with their severity and duration.

6.1 — P1 Critical Incidents

• T+0 min: Automated monitoring detects pipeline failure and pages the on-call Site Reliability Engineer (SRE).
• T+15 min: If not acknowledged, incident automatically escalates to the SRE team lead.
• T+30 min: If unresolved, escalates to the VP of Engineering with bridge call initiated.
• T+60 min: If unresolved, escalates to the Chief Technology Officer. Affected Sovereign Infrastructure customers are joined to the bridge.
• T+4 hours: Executive escalation to the CEO with mandatory status communication to all affected customers.

6.2 — P2 High Incidents

• Assigned to the on-call SRE within the applicable response time window.
• If unresolved within twice (2x) the initial response time, escalates to the SRE team lead.
• If unresolved within four times (4x) the initial response time, escalates to the VP of Engineering.

6.3 — P3 Medium Incidents

• Assigned to an engineering team member on the next business day.
• If unresolved within five (5) business days, escalates to the engineering team lead.

6.4 — P4 Low Incidents

• Tracked in the standard engineering backlog and addressed in the normal development cycle.

Service Credits

If EVE AI Core fails to meet the Monthly Uptime Percentage commitment for the applicable tier, Customer is eligible to receive Service Credits as described below. Service Credits are calculated as a percentage of the monthly fees actually paid by Customer for the affected service during the month in which the SLA was not met.

7.1 — Credit Request Procedure

1. Customer must submit a Service Credit request within thirty (30) days of the end of the calendar month in which the Downtime occurred.
2. Requests must be submitted via email to [email protected] and must include the dates and times of each Downtime event, the Customer account identifier, and applicable server logs or monitoring data corroborating the claim.
3. EVE AI Core will evaluate the request against internal monitoring data and respond within fifteen (15) business days.
4. Approved Service Credits will be applied as a credit against future invoices. Service Credits are not redeemable for cash and may not be transferred.

7.2 — Maximum Credit

The aggregate maximum Service Credit for any single calendar month shall not exceed fifty percent (50%) of the monthly fees paid by Customer for the affected service during that month. Service Credits constitute Customer’s sole and exclusive remedy for EVE AI Core’s failure to meet the SLA.

Exclusions

This SLA does not apply to any performance issues or unavailability caused by:

1. Scheduled Maintenance. Planned maintenance windows communicated at least seventy-two (72) hours in advance via the status page, email notification, or the Customer’s designated Slack channel. Sovereign Infrastructure customers may negotiate custom maintenance windows.
2. Force Majeure. Events beyond EVE AI Core’s reasonable control, including but not limited to: natural disasters, acts of war or terrorism, government actions, pandemics, widespread internet outages, or failures of third-party cloud infrastructure providers (AWS, GCP, Azure) affecting multiple independent services.
3. Customer-Caused Issues. Downtime or degradation resulting from Customer actions, including: exceeding contracted rate limits, sending malformed requests, misconfiguring API integration, or failing to implement recommended security practices.
4. Abuse or Security Incidents. Service unavailability resulting from DDoS attacks, security incidents, or actions taken by EVE AI Core to protect the integrity of the platform or its customers.
5. Beta or Preview Features. Services, endpoints, or features explicitly designated as “beta,” “preview,” or “experimental” are not covered by this SLA.
6. Free Tier Usage. This SLA applies only to paid, commercially licensed deployments. Usage under free, trial, or evaluation licenses is provided on a best-effort basis without uptime commitments.

Monitoring & Transparency

EVE AI Core maintains comprehensive monitoring infrastructure to measure, record, and report on SLA compliance.

9.1 — How Uptime Is Measured

• Synthetic monitoring: Automated health checks execute governance pipeline requests every ten (10) seconds from multiple geographic regions. A health check failure is recorded when the pipeline fails to return a valid verdict within 30 seconds.
• Internal instrumentation: Server-side telemetry records every governance request with nanosecond-precision timestamps, verdict outcome, and latency at each pipeline stage.
• Third-party validation: An independent monitoring service performs external availability checks at one-minute intervals. This data is used as an impartial reference for SLA credit disputes.

9.2 — Status Page

Real-time and historical availability data is published at status.eveaicore.com. The status page displays:

• Current operational status of all major components (Authority Resolution Layer, Governed Inference Layer, Cryptographic Authority Chain Layer);
• Active and recent incidents with severity classification and current status;
• 90-day uptime history for each component;
• Scheduled maintenance windows; and
• Subscription options for email, Slack, and webhook incident notifications.

9.3 — Customer-Accessible Metrics

Licensed customers can access their own latency and availability data through the following programmatic interfaces:

• GET /api/latency/breakdown — Per-stage p50/p95/p99 latency statistics
• GET /api/latency/waterfall — Per-request latency waterfall for recent requests
• GET /api/health/status — Current system health score and component status
• GET /api/resilience/score — Composite resilience score (0–100) with sub-layer breakdown
• GET /api/resilience/certificate — Cryptographically signed resilience certificate for audit purposes

General Terms

10.1 — Modifications

EVE AI Core may modify this SLA from time to time. Any material changes will be communicated to affected Customers at least thirty (30) days prior to the effective date of the modification. Continued use of the service after the effective date constitutes acceptance of the modified SLA. Modifications will not apply retroactively to any SLA credit claims submitted prior to the modification date.

10.2 — Conflict with License Agreement

In the event of a conflict between this SLA and the Customer’s executed license agreement, the terms of the executed license agreement shall prevail. Custom SLA terms negotiated in an Order Form or Statement of Work supersede this document.

10.3 — Governing Law

This SLA is governed by the laws of the State of Georgia, United States, without regard to its conflict of law provisions.

10.4 — Contact

For SLA-related inquiries, service credit requests, or to report an incident:

• Email: [email protected]
• Status Page: status.eveaicore.com
• Sovereign Infrastructure Emergency Line: Available to Sovereign tier customers via their dedicated support channel.