Build a proposal queue, not an autonomous executive
A safe command center collects permitted evidence, drafts a traceable proposal, and waits. Scheduling does not create authority.
The learning workflow must not send messages, publish content, change ads or prices, alter orders, move money, modify customer or employee records, execute code, or approve decisions. Any future action path needs separate authorization, least privilege, environment isolation, qualified review, evaluations, human approval, and tested rollback.
Any production implementation additionally requires approved redacted inputs, rights and lawful basis, data minimization, least privilege, tenant separation, managed secrets, audit logs, representative, adversarial, and regression evaluations, proposal-only behavior, human approval, monitoring, a stop condition, an incident path, and tested rollback.
A controlled reference architecture
Security decisions are part of the product
- Use a managed secret store; never place credentials in prompts, source control, URLs, client JavaScript, or model-visible logs
- Create separate identities per environment and purpose, with explicit read scopes and tested expiration or revocation
- Enforce tenant keys at storage and query boundaries; do not rely on a prompt to prevent cross-tenant access
- Allowlist egress and tools; validate schemas, file types, redirects, URLs, output destinations, and response sizes
- Assume documents, email, web pages, and tool output can contain prompt injection; never let content grant itself permissions
- Minimize and redact data before model access; define retention, deletion, incident notification, and provider change review
A fictional executive brief job reads a synthetic warehouse view and a curated public-source feed. The generator cannot call outbound tools. Each claim links to a source snapshot; unsupported claims fail the evaluation. The draft enters a review queue. A reviewer can publish internally, request correction, or stop the job. The source and model versions remain in the audit record.
Deploy by evidence
Define the output contract
Workflow launch gate
Use OWASP’s LLM application guidance to seed abuse cases and the NIST AI RMF to structure governance. Validate controls against the actual model, tools, identities, data, and deployment.