Built on Open Standards
Every protocol choice is deliberate — designed for interoperability, partner integration, and long-term defensibility. No proprietary lock-in.
Core Standards
OpenID AuthZEN 1.0
Policy evaluation API for interoperable authorization. EmpowerNow's PDP implements AuthZEN with constraints, obligations, and TTL extensions. Single PDP surface for LLMs and tools.
Learn more →Model Context Protocol (MCP)
Anthropic's protocol for AI agent tool integration. EmpowerNow natively publishes governed MCP tools from any connector action, with policy-scoped discovery, schema pinning, and runtime authorization.
Learn more →OAuth 2.1
Modern OAuth with mandatory PKCE and refresh rotation. EmpowerNow's credential isolation builds on OAuth 2.1 for three-zone trust boundaries.
Learn more →OAuth 2.0 DPoP (RFC 9449)
Demonstration of Proof-of-Possession binds tokens to the client that requested them. JKT binding on delegation edges prevents token theft and replay.
Learn more →Rich Authorization Requests (RFC 9396)
Fine-grained capability expression beyond OAuth scopes. Enables per-operation constraints — budget caps, velocity limits, time windows — in authorization requests.
Learn more →Token Exchange (RFC 8693)
Secure delegation chains for agent-on-behalf-of-user flows. Enables user → agent → tool identity chaining with verifiable delegation.
Learn more →SCIM 2.0
Standard provisioning protocol. EmpowerNow's SAP IAS connector implements full SCIM 2.0 with 70 MCP-enabled tools for user, group, and entitlement management.
Learn more →HTTP Conditional Requests (RFC 9110)
Workflow state versioning via ETag/If-Match. WAITING contracts use 412 Precondition Failed on mismatch for optimistic concurrency control.
Why standards matter
Architecture decisions that protect your investment
Partner-Deployable
Standards-based architecture means partners (Deloitte, Accenture, SIs) can deploy and operate without proprietary training.
OEM-Embeddable
AuthZEN-compliant PDP and MCP-native tools can be embedded into partner platforms with clean integration boundaries.
Future-Proof
Open standards outlast vendor lock-in. EmpowerNow's architecture is built to evolve with the standards ecosystem, not against it.
OWASP LLM Top-10 Mapping
How ARIA maps controls to each documented LLM risk
| OWASP Risk | ARIA Control | Component | Residual Risk |
|---|---|---|---|
| LLM01: Prompt Injection | Egress allowlist, classification policies | ARIA Shield | Advanced jailbreaks may evade |
| LLM02: Insecure Output Handling | Data scope row filtering, redaction (x-redact) | Backend + Policy | Complex output requires review |
| LLM04: Model Denial of Service | Streaming caps, rate limits, HTTP 402 | ARIA Shield | Distributed attacks require upstream |
| LLM05: Supply Chain | Schema pins (Pin JWS), MCP server verification | MCP Gateway + Edge/Signer | Compromised upstream servers |
| LLM06: Sensitive Info Disclosure | Prompt redaction, data_scope, receipts | ARIA Shield + Policy | Inference attacks on outputs |
| LLM07: Insecure Plugin Design | Parameter allowlists, egress pinning, schema validation | MCP Gateway | Tool implementation vulnerabilities |
| LLM08: Excessive Agency | Capability limits, plan discipline, delegation model | Delegation + MCP Gateway | Within-scope misuse |
| LLM09: Overreliance | User-in-loop for high-risk, approval obligations | PDP + Policy | Automation bias |
LLM03 (Training Data Poisoning) and LLM10 (Model Theft) are provider responsibilities and out of ARIA's scope.
MITRE ATLAS Mapping
AI-specific threat framework coverage
| ATLAS Tactic | ARIA Defense | Enforcement Layer |
|---|---|---|
| Resource Hijacking | Budget caps, rate limits, HTTP 402 enforcement | PDP + ARIA Shield |
| ML Model Access | Delegation + per-operation capabilities | Membership Graph |
| Evade ML Model | Schema pins prevent tool drift | MCP Gateway |
| Exfiltration via AI | Egress allowlist, data scope row filtering | ARIA Shield + Policy |
| Impact / Abuse | Signed receipt chains (tamper-evident audit) | Receipt Vault |
Selected ATLAS tactics relevant to agent-based AI systems. Full ATLAS coverage requires additional application-layer and provider-side controls.
EU AI Act Compliance
Regulation 2024/1689 — GPAI provider obligations entered force August 2025; full enforcement from August 2026
| AI Act Requirement | ARIA Implementation | Component |
|---|---|---|
| Transparency (Article 53) | Policy snapshots in receipts, SSC evidence digests | Receipt Vault, Approval Engine |
| Technical Documentation | Tool schema versioning, MCP catalog, param_schema | Edge/Signer, Membership |
| Traceability of Decisions | Signed receipts with decision_id, policy_hash, workflow fingerprints | Receipt Vault, PDP |
| Logging & Retention (Annex IV) | Tamper-evident receipt chains, WORM storage (S3 Object Lock) | Receipt Vault, Analytics |
| Risk Management (Article 9) | Classification, constraints, dual PEP enforcement, SSC risk banding | ARIA Shield, MCP Gateway, PDP |
| Human Oversight (Article 14) | Approval obligations, consent flows, requires_human enforcement | PDP, Approval Engine |
| Accuracy & Robustness | Schema pins prevent drift, parameter validation | MCP Gateway, Edge/Signer |
| Record Keeping (Annex IV) | Receipts with 6+ year retention option | Analytics + S3 Object Lock |
| Audit Trail Integrity | Per-agent hash chains, state versioning, idempotency tracking | Receipt Vault, Workflow API |
ARIA does not claim blanket AI Act compliance. Obligations vary by role (provider, deployer, integrator) and risk classification. Legal review remains necessary.
Threat-to-Control Reference
Quick lookup for security architects
Residual Risks
ARIA mitigates authorization and governance risks. The following remain provider or application responsibilities: model training security, advanced prompt jailbreak defense, tool/API implementation vulnerabilities, distributed denial-of-service attacks, and inference attacks on model outputs.
Deploy ARIA as part of defense-in-depth alongside provider-side protections (Azure Content Safety, OpenAI Moderations) and application security.
U.S. patent applications pending
All filed May 2025. Patent pending.
Related pages
Go deeper on the architecture behind these standards
See standards-based governance in action
Walk through AuthZEN authorization, MCP tool governance, and cryptographic receipts on your environment.