VORIONVORION
BASIS/NIST Submissions
Public Policy Engagement

NIST Submissions

Vorion actively participates in NIST's AI standards development. Our submissions draw from direct experience operating the BASIS-governed AI platform to provide concrete, implementable recommendations to policymakers.

RFI Response Submitted

NIST CAISI: Security Considerations for AI Agents

Vorion's response to NIST CAISI's RFI on AI agent security, drawing from direct experience building the Vorion Governed AI Execution Platform. Addresses all five RFI topics with concrete implementation patterns, quantitative data, and open-source reference implementations.

Docket

NIST-2025-0035

Submitted

Submitted March 2026

Document

Federal Register Vol. 91, No. 5 (pp. 698-701)

Threat Model

10 threat categories specific to agentic AI, mapped to OWASP Top 10 for Agentic Applications

Control Pattern

Trust-tiered gating: T0-T7 tiers with 16 behavioral factors governing what agents can do

Open Source

Full reference implementation at github.com/vorionsys/vorion (Apache-2.0)

Topic 1

Security Threats, Risks & Vulnerabilities

Traditional software threats exploit implementation bugs. Agent threats exploit the fundamental architecture -- agents receive instructions in the same medium they process data (natural language), operate with persistent state, and combine tools in emergent ways no individual tool author anticipated.

10 Agentic Threat Categories (OWASP Mapped)

OWASPThreatWhy Agents Are Different
ASI01Goal hijackingNatural-language instructions cannot be distinguished from data; prompt injection exploits meaning, not syntax.
ASI02Tool weaponizationAgents chain legitimate tools in unintended sequences; individually safe tools create attack paths when combined.
ASI03Identity inheritanceAgents inherit human-level permissions by default with no established "agent identity" pattern.
ASI04Supply chain compromiseMCP servers, plugins, and prompt templates are loaded dynamically from unverified sources at runtime.
ASI05Code execution escapeAgents generate and execute code as normal operation; the boundary between "output" and "command" is blurred.
ASI06Memory poisoningA single successful injection persists across sessions; every future interaction inherits the compromise.
ASI07Inter-agent spoofingMulti-agent communication uses natural language with implicit trust -- no TLS equivalent exists.
ASI08Cascading failuresConnected agent systems amplify errors exponentially; one compromised agent poisons downstream chains within hours.
ASI09Trust exploitationAgents generate authoritative explanations that turn human-in-the-loop review into rubber stamps.
ASI10Rogue behaviorAgents may develop misaligned objectives through reward hacking or memory drift without any external attacker.

Barriers to Adoption (Question 1c)

01
Liability uncertainty

Without graduated trust and containment, the risk profile of deploying an autonomous agent is binary -- it works or it causes damage with no intermediate states.

02
Audit gap

SOC 2 and ISO 27001 have no controls specific to AI agent behavior. Regulated industries cannot demonstrate compliance for agent deployments.

03
Insurance gap

Cyber insurance policies typically exclude AI incidents or lack underwriting models for agentic risk. Without quantifiable trust metrics, insurers cannot price the risk.

Multi-Agent Specific Threats (Question 1e)

Trust chain propagation

When Agent A trusts Agent B's output for downstream decisions, a single compromise propagates through the chain. There is no "certificate revocation" for poisoned outputs already consumed.

Emergent behavior

Agents may each operate within policy boundaries, yet their collective behavior produces outcomes no single agent's policy was designed to prevent -- the composition problem applied to natural language.

Authority escalation via delegation

A low-trust agent may request a higher-trust agent to perform actions it cannot do directly. Without explicit delegation controls, multi-agent systems bypass per-agent access controls.

Topic 2

Security Practices: Mitigations & Technical Controls

Vorion recommends three categories of controls informed by operational experience: model-level robustness controls, agent system-level graduated trust architecture, and cryptographic proof chains.

Graduated Trust Architecture -- T0 through T7

TierScoreAgent CapabilitiesGovernance Posture
T0 Sandbox0-199Read-only, no external callsAll intents require approval
T1 Observed200-349Basic tools, scoped dataEnhanced logging active
T2 Provisional350-499Standard tools, rate-limitedSensitive ops require review
T3 Monitored500-649Full standard toolsetContinuous monitoring
T4 Standard650-799Extended tools + external APIsGreen-light for most operations
T5 Trusted800-875Cross-namespace accessElevated authority scope
T6 Certified876-950Administrative operationsCan approve others' intents
T7 Autonomous951-1000Unrestricted within policySelf-governing

Scoring Design Principles

Start at zero

All new agents begin at T0 (Sandbox). Trust is earned, never assumed.

Asymmetric scoring

Failures penalize trust more heavily than successes reward it -- a deliberate design bias toward reliability over speed of trust acquisition.

Stepped decay

182-day (6-month) half-life via 9 milestones. Grace period: days 0-6. Day 7: -6%, Day 14: -12%, Day 28: -18%, Day 56: -30%, Day 112: -40%, Day 182: -50%. Any activity before a milestone resets the clock.

Behavioral signals

Scores computed from: Behavioral (40%), Compliance (25%), Identity (20%), Context (15%).

Fluid Governance -- Three-Tier Decision Model

GREEN -- Approved

Proceed with constraints: allowed tools, data scopes, rate limits, execution time, and reversibility requirements.

YELLOW -- Refine

Transform “access denied” into collaborative negotiation. Agent can reduce scope, add constraints, request approval, or decompose the intent.

RED -- Denied

Hard policy violation. Triggers containment escalation and trust score decay. No negotiation path.

Cryptographic Proof Chains

Dual-hash chain

SHA-256 links each record to its predecessor; parallel SHA3-256 integrity anchors provide algorithmic diversity and a migration path if either is weakened (ADR-017).

Digital signatures

Ed25519 signatures bind each record to a specific agent identity -- non-repudiation for every governance decision.

Merkle aggregation (planned)

Periodic Merkle tree construction is scaffolded for external anchoring and batch verification in high-frequency deployments.

Privacy-preserving verification (planned)

Pedersen commitment and range proof interfaces are scaffolded to enable agents to prove trust tier membership without revealing exact scores. Production integration with ristretto255/circom is planned.

Reference implementation maturity: 400+ unit tests for the trust engine · 97 automated NIST SP 800-53 Rev 5 control verification tests · Continuous SBOM (CycloneDX 1.5 + SPDX 2.3) · Machine-readable OSCAL artifacts (SSP, component definition, POA&M) · ENFORCE gate <100ms (typical ~30ms) · PROOF logging <50ms (typical ~10ms)
Topic 3

Assessing Security

Assessment spans four dimensions: trust posture, capability scope, governance coverage, and proof chain health. These build on established SIEM and behavioral analytics practices, adapted for the unique characteristics of AI agents.

Four Assessment Dimensions

1
Trust posture

Current trust tier, score history, and trajectory. An agent with a stable T4 score and no recent containment events presents lower risk than an agent oscillating between T2 and T4.

2
Capability scope

What tools, data, and external services can the agent access? Assessment must include both explicitly granted and implicitly available capabilities (tools accessible through other tools).

3
Governance coverage

What percentage of possible actions are covered by explicit policy rules? Systems with gaps in P0--P2 coverage present unacceptable risk.

4
Proof chain health

Chain completeness, signature validity, chain integrity verification pass rate, and anchoring freshness -- monitored continuously.

Governance Coverage -- Policy Priority Levels

PriorityCategoryPurposeAssessment Metric
P0Hard disqualifierImmediate denialFalse negative rate (target: 0%)
P1Regulatory mandateComplianceCompliance gap %
P2Security criticalViolation detectionDetection latency (ms)
P3Policy enforcementStandard policiesOverride rate
P4Soft constraintGuidelinesAcknowledgment rate
P5Clarification triggerAmbiguity resolutionClarification success rate
P6Logging onlyAudit trailCoverage %

Post-Deployment Incident Detection Methods

Trust score monitoring

Sudden drops indicate behavioral anomalies. Default: 100-point drop triggers containment escalation. Configurable per deployment.

Proof chain analysis

Continuous verification detects tampering or audit trail gaps. Chain completeness, signature validity, and hash integrity are security metrics.

Behavioral drift detection

Compare tool usage frequency, error rates, and operation scope against established baselines. Significant deviations trigger enhanced monitoring.

Escalation frequency monitoring

Agents repeatedly triggering YELLOW or RED decisions indicate policy misconfiguration or genuine behavioral instability -- both require investigation.

Topic 4

Limiting, Modifying & Monitoring Agent Access

Progressive containment provides proportional response to detected threats across 7 levels. Every GREEN governance decision attaches specific constraints -- allowed tools, data scopes, rate limits, max execution time, and reversibility requirements.

7-Level Progressive Containment

LevelNameRestrictionsTrigger
L1Full AutonomyNoneDefault for T5+ agents
L2MonitoredEnhanced loggingTrust score below T4 threshold
L3Tool RestrictedHigh-risk capabilities blockedPolicy violation detected
L4Human-in-LoopAll actions require approvalTrust score drop >100 points
L5Simulation OnlyRead-only, no side effectsSuspected goal hijacking
L6Read OnlyWrite/execute blockedConfirmed anomalous behavior
L7HaltedALL operations blockedKill switch activated

Zero-Trust Adaptations for Agents

Continuous identity verification

Agent identity verified on every request, not just at session establishment. Trust scores re-evaluated continuously -- not just at authentication time.

No default trust

No agent is trusted by default, regardless of its creator or deployment context. All agents start at T0.

CAR-ID agent identity

Categorical Agentic Registry provides unique, verifiable identity per agent distinct from human credentials. No credential inheritance.

Adversarial inter-agent posture

All agent-to-agent communication is treated as potentially adversarial. Trust scores propagate -- agents do not blindly accept instructions from lower-trust agents.

Legal & Privacy Considerations

Data retention

Proof chains may contain sensitive data (user queries, agent reasoning traces). Retention policies must balance auditability with privacy requirements.

GDPR right-to-erasure

Conflicts with immutable proof chains. Addressed through pseudonymization -- proof records reference entity IDs, not personal data -- and Pedersen commitment interfaces.

Employee monitoring

Monitoring agents acting on behalf of specific employees may implicate privacy regulations in some jurisdictions. Clear scope and purpose policies are required.

Topic 5

Additional Considerations

Open-source reference implementations paired with open standards are the most effective adoption mechanism. The U.S. should lead on runtime agent governance standards -- the critical gap between pre-deployment evaluation (where other jurisdictions have invested) and runtime behavioral controls (where no jurisdiction has published guidance).

Critical Research Priorities (Question 5c)

01
Prompt injection defenses

Despite 3+ years of identification, no robust defense exists. Research should focus on architectural solutions -- separating instruction and data channels -- rather than input filtering.

02
Multi-agent trust propagation

Formal models for how trust should propagate, decay, and revoke across agent fleets. Current approaches are entirely ad hoc.

03
Behavioral drift detection

Methods to detect when an agent's behavior has shifted from its intended purpose through subtle changes that individually appear benign but collectively represent goal drift.

04
Agent-to-agent authentication

Cryptographic protocols for agent-to-agent communication providing authentication, integrity, and non-repudiation without prohibitive latency.

05
Containment effectiveness measurement

Empirical research on how effectively containment mechanisms prevent harm propagation, and what levels are appropriate for different risk categories.

Most Urgent Government Collaboration Areas (Question 5b)

Agent identity standards

No established standard exists for AI agent identity distinct from human identity or service accounts. Government collaboration analogous to PIV/CAC for humans would address ASI03, ASI07, and ASI09 simultaneously.

Multi-agent trust propagation

As agencies deploy multi-agent systems, trust propagation between agencies' agents will require interoperable trust frameworks. NIST is uniquely positioned to develop cross-organizational agent trust standards.

Incident reporting channels

AI agent security incidents are currently unreportable through existing channels (CISA, CVE). A mechanism for prompt injection campaigns, supply chain compromises, and behavioral drift events would improve collective defense.

Compliance framework alignment

FedRAMP and FISMA have no mapping to AI agent security controls. NIST guidance on how agent security controls satisfy existing compliance requirements would accelerate safe adoption.

Applicable Practices from Other Fields (Question 5e)

Aviation safety

Crew resource management -- multi-layered authority model (captain, first officer, ATC) maps to trust tiers and escalation chains.

Nuclear safety

Defense in depth -- multiple independent barriers principle informs progressive containment design. No single failure should cascade.

Financial services

Graduated authority limits based on demonstrated competence and track record. Trust scoring adapts this for AI agents.

Pharma manufacturing

FDA process validation (IQ/OQ/PQ) maps to agent lifecycle: installation, operational, and performance qualification.

Industrial control systems

SCADA safety systems are physically separate from control systems. Agent kill switches must be architecturally isolated from the agent control plane.

Alignment

Standards Alignment

StandardHow This Work Aligns
NIST AI RMF (AI 100-1)GOVERN: Trust tiers and policy rules | MAP: Threat taxonomy | MEASURE: Trust scoring metrics | MANAGE: Progressive containment and kill switches
NIST CSF 2.0Extends ID, PR, DE, RS, RC functions with agent-specific controls; agents addressed as a distinct asset type
NIST IR 8596 (Cyber AI Profile)Builds on our January 2026 public comment prepared for NIST's Cyber AI Profile
OWASP Top 10 for Agentic AIFull ASI01-ASI10 mapping with implemented technical controls per threat category
ISO/IEC 42001Trust scoring complements AIMS with runtime behavioral measurement
EU AI ActTrust tiers map to risk categories; progressive containment addresses Article 14 human oversight requirements

Vorion's Open Standards Work

The BASIS standard and Vorion platform are fully open-source. All governance infrastructure referenced in these submissions is available for NIST review and public implementation.