The Core Argument
Every lab already depends on
training data vendors.
The question is not whether to use them. The question is whether you shape the next one from the foundation — or audit its failures after they occur. Mercor was built without input from the labs it served. Its security architecture, labor model, and pricing structure were designed to serve Mercor's growth. The result was a platform that became a single point of failure for the entire AI training supply chain.
Co-pilot status is the structural alternative. Your security requirements become the architecture. Your evaluation standards become the platform default. Your compliance specifications are written into the system before a line of production code is deployed.
Architecture Approval Rights
Formal sign-off required on security architecture and client isolation design before production deployment. Your security team reviews the architecture before it is built — not after it fails.
Pricing Locked for 3 Years
20-22% take rate locked by contract for three years regardless of platform growth or market pricing changes. At $50M annual volume, this saves $6-7.5M per year versus current market rates.
Equity Stake + Board Observer
0.5-1.5% equity stake (depending on contract size) provides economic upside as the platform scales. One non-voting board observer seat gives visibility into strategic decisions.
Unannounced Audit Rights
Any co-pilot lab may trigger an unannounced audit of security posture, expert credential compliance, and data handling at any time. No advance notice required. Results delivered within 5 business days.
Priority Onboarding
Co-pilot labs are onboarded first. Expert pools, rubric frameworks, and task pipelines are built to your specifications before the platform accepts any non-co-pilot clients.
10 Mandatory Requirements
Each co-pilot lab submits up to 10 specific technical requirements or pain points at contract signing. These are written into the platform's technical specifications as mandatory implementation requirements.
Term Sheet Summary
What every co-pilot lab
receives.
| Term | Specification |
|---|---|
| Take Rate | 20% under $20M annual volume · 19% over $20M · 18% over $50M — locked for 3 years regardless of market changes. |
| Equity Stake | 0.5% — $5M pilot · 1.0% — $10M annual contract · 1.5% — $20M+ annual contract. Standard preferred stock, pro-rata rights on future rounds. |
| Board Observer | One non-voting observer seat. Receives all board materials. May attend all board meetings. Cannot vote or exercise board authority. |
| Architecture Approval | Security architecture and client isolation design require formal co-pilot lab approval before production deployment. Veto rights limited to security architecture only. |
| Audit Rights | Unannounced right to audit security posture, expert compliance, and data handling at any time. Results delivered within 5 business days. Cost borne by Provenance AI. |
| Mandatory Requirements | Up to 10 technical requirements per lab written into platform specifications as mandatory implementation items. Failure to implement triggers contract breach remedy. |
| Priority Onboarding | Co-pilot labs onboarded before any non-co-pilot client. Expert pool credentialing and rubric framework built to your specifications during the pilot period. |
| Exit Rights | 12-month notice exit right if the platform fails to meet agreed security or quality standards. Exit triggers equity buyback at fair market value. |
| Contract Term | 12-month initial term with automatic renewal. Quarterly payment schedule. Volume discounts applied at each anniversary based on trailing 12-month volume. |
Lab Advisory Council
How governance works without
creating conflicts.
The most common question about the co-pilot model: if Anthropic and Google DeepMind both have governance participation, won't they learn things about each other? The answer is no — and the architecture that makes that guarantee enforceable is documented below.
Lab Advisory Council — Operating Structure
Full Council Scope
- Security architecture standards
- Platform-wide quality minimums
- Incident response protocols
- Roadmap input (non-confidential items only)
- Quarterly aggregate metrics (no client-specific data)
1:1 Session Scope
- Lab-specific rubric design
- Lab-specific expert pool standards
- Lab-specific compliance requirements
- Contract terms and pricing
- Lab-specific audit results
Permanently Out of Scope
- Other labs' contract terms
- Other labs' rubric content
- Other labs' data or evaluations
- Other labs' expert pool composition
- Any pricing comparison between labs
Conflict Prevention
How every conflict is prevented
structurally — not by promise.
Risk: Lab A sees Lab B's rubric methodology
Each lab's rubrics stored in isolated database schemas with row-level security. Council meetings never discuss specific client rubrics. No rubric content appears in shared council materials.
Enforcement: Cross-contamination NDA + annual third-party access control audit
Risk: Lab A learns Lab B's contract pricing
Pricing handled exclusively in 1:1 sessions between CEO and each lab. Pricing information never enters council meeting materials. CFO controls financial data access.
Enforcement: Financial data access controls audited quarterly
Risk: Security breach in Lab A's environment reaches Lab B
Completely separate cloud tenants. A security event in one tenant cannot propagate to another by architecture — not by firewall, by physical infrastructure separation.
Enforcement: SOC 2 Type II + annual isolation verification audit
Risk: Lab uses governance position to slow-walk a competitor
Veto rights are limited to security architecture changes only. Operating Agreement explicitly prohibits using governance participation to delay or block another lab's onboarding or operations.
Enforcement: Operating Agreement + legal remedy for documented misuse
The 90-day window is
open right now.
Co-pilot status costs 90 days of engagement. Waiting costs the next breach. Request a briefing — 30 minutes, direct conversation with the founder, no procurement process required.