AI Usage & Disclosure Policy
How HookPilot's 40-department AI workforce operates — and how we keep its output trustworthy.
1. What HookPilot Is
HookPilot is an AI Workforce Operating System. We don't sell a single chat box — we sell a structured organization of AI agents that plan, draft, review, and ship work alongside your human team. Behind the simple product surface is a governance layer that includes department charters, named agents at six qualification tiers, framework libraries, hallucination self-check, evidence and policy memory, output confidence scoring, escalation routing, and veto-holding agents that can stop unsafe or non-compliant work from shipping.
2. Departments
HookPilot operates 40 chartered AI departments. Each department has a published mandate, a framework set, and a list of agents:
3. Tiered Agents
Each department staffs agents at six qualification tiers: Assistant → Operator → Analyst → Specialist → Department Lead → Executive Advisor. Tier governs scope of authority, the workflows the agent may run independently, the templates it may sign off on, and the cost/model class the routing layer is allowed to spend.
4. Independent Veto Authority
Certain roles — typically the Chief Officer, Statute (compliance), Shield (security), and Guard (safety) agents in a department — hold independent veto authority. A veto stops a workflow from shipping. Vetos are recorded, attributed, and visible in the audit trail. You can review and appeal a veto, but you cannot remove the role of the veto itself; that is part of the platform's safety design.
5. Hallucination Self-Check
Every customer-visible output is run through a hallucination self-check that compares the draft against (a) the prompt and supplied facts, (b) the relevant framework library, and (c) agent memory. When a claim cannot be supported, the platform either marks the claim with a warning, asks the agent to re-draft with evidence, or escalates to a higher-tier reviewer.
6. Human-in-the-Loop
For workflows that touch the real world — legal filings, medical content, financial advice, regulated communications, hiring decisions, broadcast publications, contract signatures, money movements — HookPilot expects a human reviewer (you, your teammate, or an enterprise-designated approver) to clear the output before it ships. The product surfaces a review queue, a confidence score, and an attestation prompt so the human reviewer can sign off knowingly.
7. Mata v Avianca Discipline
Following well-publicized incidents where unchecked AI output reached courts and customers with fabricated citations and false facts, HookPilot operates under what we call Mata v Avianca discipline: never let unreviewed AI output be presented as authoritative in a forum where authority matters. This is a structural commitment, not a marketing line — it is encoded in our agent charters, in our review queues, and in the veto-holding roles described above.
8. Model Routing and Transparency
HookPilot routes requests across multiple AI providers (including Anthropic, OpenAI, Google, and open-weight models). The routing layer chooses a model based on the task, the customer's plan, cost budget, latency requirements, and any compliance constraints (for example, dual-review across providers for Tier 4 regulated outputs). Enterprise customers can request workspace-level model restrictions. We do not enable training on Customer Content; see Privacy Policy §6.
9. What HookPilot Does Not Do
HookPilot does not promise viral reach, revenue, rankings, regulatory approval, or any specific business outcome. HookPilot does not provide professional legal advice, medical diagnosis, individualized financial advice, or licensed engineering sign-off. Where a domain requires a licensed professional, HookPilot helps that professional work faster — it does not replace them.
10. Your Responsibilities
You agree to (a) review AI output before publication or commercial use, (b) keep a human in the loop for any decision Section 6 would call material, (c) avoid prompting HookPilot to produce content prohibited by our Acceptable Use rules, (d) honor third-party rights when uploading source material, and (e) clearly disclose AI-assisted content where the platform you publish to requires disclosure.
11. Reporting Bad Output
If HookPilot produces output that is wrong, harmful, or violates this policy, please report it via the in-product "Report this output" control or email support@hookpilot.co. Severe issues (safety, defamation, IP infringement) are routed to legal@hookpilot.co and the platform's incident response runbook.
12. Contact
AI policy questions: legal@hookpilot.co
Product feedback: support@hookpilot.co