Examiners from the OCC, Fed, and SEC don't accept 'the model decided' as an answer. AutoPIL enforces access policy at the data layer and produces a tamper-evident audit trail that maps every AI decision to the governing policy — automatically.
25 policies across consumer banking, fraud investigation, wealth management, risk & compliance, and operations — built for the strict data boundary requirements of retail and institutional finance.
policies: # Loan underwriter — credit and collateral data; blocked from peer data and internal models - name: loan_underwriter_policy agent_role: loan_underwriter allowed_sources: - account_summaries - credit_scores - loan_history - property_valuations - income_verification denied_sources: - other_customer_data - internal_risk_models - executive_communications allowed_tasks: - credit_decision - collateral_check - risk_assessment denied_tasks: - account_freeze - fraud_flag - product_recommendation max_sensitivity: high
Point policy_path at your industry directory. AutoPIL loads every
YAML file recursively — roles, sensitivity rules, and process groups wired up
automatically from the directory structure. No additional config needed.
Policies are hot-reloaded at runtime. Extend or override any pre-built policy via the REST API without redeploying your agents or restarting services.
from autopil import ContextGuard # Point at your industry — only those policies load guard = ContextGuard( policy_path="policies/financial_services/", audit_db="autopil.db", ) # policies/financial_services/ — loads recursively # Switch verticals by changing the path — nothing else changes.
Pre-built policies for financial services are included in every AutoPIL trial. Extend or override any rule via the REST API without redeploying.