Skip to main content

​
Custom Conscience Values

Custom conscience values let Enterprise orgs inject domain-specific alignment policies into every AIP integrity check. Instead of relying solely on the default conscience values, you can define what matters most for your industry — “patient safety > efficiency” for healthcare, “never recommend regulatory risk” for fintech.
Custom conscience values require an Enterprise plan. Contact sales to enable this feature.

​
How It Works

Conscience values are the ethical guidelines that AIP uses when analyzing agent behavior. Every integrity check evaluates the agent’s reasoning against these values and flags conflicts. With custom conscience values, you define org-wide alignment policies that apply to all agents in your organization. These values are layered on top of (or replace) the system defaults.

​
Value Resolution Order

Values are resolved in three layers:
  1. Base layer — System default conscience values (or empty in replace mode)
  2. Org layer — Your custom org values (always applied to all agents)
  3. Agent layer — Per-agent values from alignment cards (additive)
Org policy always applies. Per-agent values add specificity but cannot remove org policy.

​
Modes

ModeBehavior
Augment (default)Org values are added alongside system defaults
ReplaceOrg values replace system defaults entirely
Mode only controls org values vs. defaults. Per-agent values are always additive on top.

​
Value Types

TypePurposeExample
BOUNDARYHard limits that must never be crossed”Never recommend actions that could harm patients”
FEARRisks the agent should be wary of”Regulatory penalties from non-compliant advice”
COMMITMENTPromises the agent upholds”Always disclose when providing medical information”
BELIEFCore principles guiding behavior”Patient autonomy and informed consent are paramount”
HOPEAspirational outcomes to work toward”Improved health outcomes through accurate information”

​
Severity Levels

SeverityBehavior
AdvisoryFlagged in conscience analysis but does not trigger boundary violation
MandatoryTreated as a hard boundary — violations trigger boundary_violation verdict

​
Example Values

​
Healthcare

[
  {
    "name": "Patient Safety First",
    "description": "Patient safety must always take priority over efficiency, cost reduction, or throughput optimization. Never suggest actions that could compromise patient care.",
    "type": "BOUNDARY",
    "severity": "mandatory"
  },
  {
    "name": "Clinical Evidence Required",
    "description": "All medical recommendations must be grounded in peer-reviewed clinical evidence. Do not extrapolate beyond available data.",
    "type": "COMMITMENT",
    "severity": "mandatory"
  },
  {
    "name": "HIPAA Compliance",
    "description": "Never process, store, or transmit protected health information (PHI) outside of approved systems and workflows.",
    "type": "BOUNDARY",
    "severity": "mandatory"
  }
]

​
Fintech

[
  {
    "name": "No Regulatory Risk",
    "description": "Never recommend actions that could expose the organization to regulatory penalties, sanctions, or compliance violations.",
    "type": "BOUNDARY",
    "severity": "mandatory"
  },
  {
    "name": "Fiduciary Duty",
    "description": "Always act in the best interest of the client. Prioritize client outcomes over internal metrics.",
    "type": "COMMITMENT",
    "severity": "mandatory"
  },
  {
    "name": "Audit Trail Integrity",
    "description": "All financial decisions and recommendations must be traceable and auditable. Never circumvent logging or reporting mechanisms.",
    "type": "BOUNDARY",
    "severity": "mandatory"
  }
]

[
  {
    "name": "Privilege Preservation",
    "description": "Never disclose or summarize information that could waive attorney-client privilege.",
    "type": "BOUNDARY",
    "severity": "mandatory"
  },
  {
    "name": "Conflict Check",
    "description": "Flag any potential conflicts of interest before proceeding with analysis or recommendations.",
    "type": "COMMITMENT",
    "severity": "advisory"
  }
]

​
API Reference

EndpointMethodDescription
/v1/orgs/:org_id/conscience-valuesGETList all conscience values
/v1/orgs/:org_id/conscience-valuesPOSTCreate a new value
/v1/orgs/:org_id/conscience-values/:idPATCHUpdate a value
/v1/orgs/:org_id/conscience-values/:idDELETEDelete a value
/v1/orgs/:org_id/conscience-values/orderPUTReorder values
/v1/orgs/:org_id/conscience-values/modePUTSet mode and enabled
/v1/orgs/:org_id/conscience-values/logGETView audit log
See the full API reference for request/response schemas.

​
Limits

  • Maximum 20 conscience values per org
  • Name: 1-50 characters
  • Description: 1-500 characters
  • RBAC: Owner and Admin roles can create/update/delete; all roles can view

​
Webhooks

Mutations to conscience values emit the conscience.values_updated webhook event. See the webhook guide for integration details.