Aevesa Trust Bundle

Enterprise trust & operational sovereignty

• Verification-first governance. Tamper-evident receipts bound to authorized intent; validate offline with @aevesa/verify without relying on a vendor dashboard.
• Deployment control & data minimization. Edge-first enforcement; verification can run locally without exporting raw agent payloads or prompt text to third-party analytics. SaaS, dedicated instance, and VPC/on-prem options.
• Open evidence for reviewers. This bundle, the Open Evidence API, and liability-receipt specifications for threat models, control mappings, and fail-closed behavior review before a pilot.

Headquartered in Los Angeles, California. Engineering and protocol design are led from the United States.

Receipt-before-action invariant Phase 2

Compliance records cannot be dropped or bypassed by an exploited runtime. The Merkle permit receipt is committed before the MCP bridge allows downstream tool execution. If the ledger cannot accept the commit, the tool path fails closed (PEP_RECEIPT_COMMIT_FAILED).

Fortress PDP & autonomy ladder Phase 3

• Out-of-process PDP. Optional aevesa-pdp-sidecar: policy decisions in an isolated process an RCE-compromised app cannot rewrite (docker compose --profile fortress).
• Evidence-based autonomy. Agents earn write authority from verifiable receipt history (Observer → Autonomous tiers via GET /api/v1/governance/evidence-portfolio/:agentId).
• GDPR v2 erasure. Art. 12 packs include crypto-shred erasure certificates; Merkle chain stays intact. Fortress tier: HSM-backed DEK wrap (roadmap).

Conformance: npm run test:phase3-isolation-conformance (19 checks).

Dual-vendor architecture · Cyera + Databricks AI-SPM Phase 4

Unity AI Gateway routes. Cyera classifies and protects data in the AI runtime. Aevesa is the independent Prove layer. Gateway attest cosigns permit/deny decisions into entryHash receipts anyone can verify offline.

• Unity + Cyera + Aevesa reference architecture
• Cyera AI Guardian battlecard (sales)
• Databricks / Unity AI Gateway attest guide
• Evidence Gap demo script (90s)
• Executed-action SBOM · ASI09 one-pager (Wave 1.3)
• RFP insert: config vs behavioral inventory
• Sample agent inventory pack

Core documents (Phase 1 + 2a)

• Infrastructure, HA & fail modes
• EU AI Act prEN conformance matrix
• OWASP ASI01-10 control matrix
• MCPShield coverage matrix 70% documented
• Consolidated threat model
• PEP / PDP architecture
• Art. 14 human oversight checklist
• Receipt-before-action conformance open standard
• Open standards index open standard
• Cryptographic vs declarative evidence open standard
• receipt-before-action/v1 badge & spec
• Conformance Lab v1: open programs & interop

SLO & production posture (summary)

MCP ingress gate p50

≤ 1.5 ms (CI harness)

MCP ingress gate p99

≤ 8 ms (harness); < 50 ms end-to-end roadmap target

Fail mode default

fail_closed_regulated: high-risk MCP/tool paths block when Redis or critical deps are unavailable

Offline receipt verify

@aevesa/verify in browser, no login on verify.aevesa.com

Full index & roadmap phases

Phase 2b-4 extensions (receipt v2, path policy, witness log) are indexed in the master document.