custom

Enterprise Agentic AI Governance

Least privilege, human oversight, grounding, and emergency shutdown for autonomous agents.

Use this package in Studio →Test in simulator ← All packages

What this package does

A cross-industry agentic AI baseline aligned with NIST AI RMF and OWASP agentic application guidance.

Designed for

AI platform teams
Security teams
Risk and compliance

What's included

Approved models and tools
Irreversible-action approval
Sensitive-data and tenant isolation
Grounding, disclosure, and kill switch

Controls in this bundle

Control	Action	Parameters
`approved_model_provider`	block	`allowed_providers=["openai"], allowed_models=[]`
`agent_tool_allowlist`	block	`allowed_tools=["read"]`
`irreversible_action_approval`	review	—
`sensitive_data_egress`	block	—
`cross_tenant_leakage`	block	—
`required_grounding`	review	`minimum_citations=1`
`required_ai_disclosure`	review	—
`high_impact_human_oversight`	review	—
`incident_kill_switch`	block	—
`prohibited_use_case`	block	`prohibited_use_cases=["social_scoring","emotion_inference_employment"]`
`agent_spend_limit`	review	`max_spend_usd=100`
`required_human_escalation`	review	`minimum_confidence=0.8`
`model_version_approval`	block	—
`predeployment_evaluation_gate`	block	`required_evaluations=["security","adversarial"]`
`immutable_log`	flag	`retention_days=2555`

Profile tiers

Switch profiles in Studio to retune default thresholds across the whole bundle without rewriting any control by hand. This package ships at enterprise — anything you've already tightened by hand is preserved on switch.

Every control in this package uses identical parameters across all four profiles. Switching profile in Studio has no effect here.

Attestation

The canonical hash of these bundle bytes is sha256:0f86b1e0ee02bd5308aba2047276e869dcabe28e8ff3ea8d0162cc4a9a7580db. The same hash is computed at lock time, at Stripe checkout, and again on the runtime side before any byte is honored.

Want to customize first? Opening this package in Studio prefills the canvas with the 15 controls above. You can add, remove, or retune any of them before you lock the hash and pay.

Open in Studio

What this package does

A cross-industry agentic AI baseline aligned with NIST AI RMF and OWASP agentic application guidance.

Designed for

AI platform teams

Security teams

Risk and compliance

What's included

Approved models and tools

Irreversible-action approval

Sensitive-data and tenant isolation

Grounding, disclosure, and kill switch