# Founder AI policy referral: a clause-to-review map | SourceX

Canonical: https://refer.sourcex.si/guides/a-referral-conversation-for-a-founder-with-an-existing-ai-policy
Last updated: 2026-11-04

Treat the founder's AI policy as their company's document, not a hurdle to help them clear. Many AI policies govern how staff use AI tools; licensing the company's own records to an AI developer is a different act, and the policy usually routes it to an approval rather than answering it. Show the founder which clauses SourceX's process will meet, leave the reading to the company and its counsel, and introduce only when the CEO, not just the founder you know, wants the conversation.

Your relationship with the founder does not change who decides. Delaware's corporate statute puts a corporation's business and affairs under the direction of its board unless the statute or the certificate of incorporation says otherwise, and in this program it is the CEO who authorizes opening a licensing discussion and who signs on the company's behalf. A board seat earns a fund a hearing, not the right to start that discussion, and it never turns the founder's interest into the company's decision.

## What the founder is asking when they mention the policy

"We have an AI policy" usually hides one of three questions, and each has a different owner.

| What the founder says | The real question | Who answers it | The next step they may want |
|---|---|---|---|
| "Does our policy even allow this?" | How the company's own document applies | The company: the policy owner and counsel | A one-page note to take to the policy owner |
| "Would this break what we promised customers?" | Whether the company can license without breaching client commitments | Counsel first, then SourceX's rights check | Nothing until counsel has looked |
| "Who would have to sign off?" | Authority | The CEO, plus any approver the policy names | A call with SourceX once the CEO agrees |

None is yours to answer. Recognize which one you heard, name its owner and offer the smallest next step the founder asked for.

## Why an AI-use policy rarely settles a licensing question

Ask which way the policy points. Many AI policies treat the company as a user of AI: which tools staff may use, what they may paste into them, how AI vendors are reviewed. Licensing reverses the direction: the company becomes the source of data, under a signed agreement and approved processing routes.

Governance frameworks explain why a policy tends to describe a process rather than a verdict. NIST's AI Risk Management Framework frames its GOVERN function around policies, processes, procedures and practices for mapping, measuring and managing AI risks being in place, transparent and implemented, with legal and regulatory requirements involving AI understood, managed and documented. A policy built in that spirit tells the company how to decide a new use of its data. It seldom contains the decision itself, which is why "does the policy allow it?" usually ends with "it depends who approves it."

## The policy-to-review question map

Each row is a clause type a founder may find, where it meets SourceX's published process, and who inside the company answers it. The founder reads the clause; you only point at the row.

| Clause the founder may find | Bears on licensing? | Where it meets SourceX's process | Who at the company answers | Your move |
|---|---|---|---|---|
| Approved AI tools, and what staff may enter into them | Rarely; licensing is not tool use | Nowhere directly: records move only under a signed agreement and approved processing routes | CTO or security lead | Never describe a license as "using an AI tool" |
| Customer and user data promises in terms, privacy policy or contracts | Directly | The rights check: licensing without breaching client confidentiality or third-party contracts | Counsel, then the CEO | Note that a promise exists; never interpret it |
| Data classification levels | Directly, for scope | The inventory: which record types, years and systems are in scope | Systems or data owner | Ask metadata questions only |
| Approval route for new data uses, such as an AI committee or the board | Directly, for authority | Before any application: who must agree inside the company | The CEO, with whoever the policy names | Wait for the company's own approval |
| Personal data and privacy | Directly | The supplier confirms personal data was collected lawfully (GDPR/CCPA); data is de-identified through approved routes | Privacy lead or counsel | Never handle samples |
| Rules on AI-generated content | Directly, in an unexpected way | The supplier confirms the data was not created or altered with AI for the sale | Records owner | Mention it as a confirmation, not a hurdle |
| Existing data-sharing or licensing agreements | Directly | The supplier confirms the data has not been licensed for AI training before; licenses are exclusive for AI training | CEO and counsel | Ask only whether one exists |
| Vendor and third-party review | Possibly | SourceX, as the company's counterparty, may go through the company's vendor review | Procurement or legal | Let the company run it |
| Retention and deletion schedules | Directly, for timing | Which years of records still exist when the inventory is taken | Records owner | Note documented dates; invent no deadline |

Read the map for its shape: one row barely matters, most route away from the founder, and one can end the conversation.

## The customer-promise row can end the conversation, and that is fine

If a company has told its customers their data will not be used to train AI, that promise outranks any enthusiasm in the room. The Federal Trade Commission's technology staff warned in February 2024 that it may be unfair or deceptive for a company to start using consumer data for AI training and announce it only through a quiet, retroactive change to its terms of service or privacy policy. The post is staff guidance, not a rule, but it marks the line clearly enough for a referral conversation.

What that means in practice is for counsel to say. A promise may cover customer content and leave the company's own internal records, such as procedures, QA notes and project histories, outside it. It may cover everything. SourceX's own screening already treats material that belongs to clients as usually not a fit. For the wider debate, see [the terms-of-service backlash over AI training](/guides/terms-of-service-ai-training-backlash) and [how SaaS terms restrict AI training data](/questions/saas-terms-of-service-ai-training-data-restrictions).

## Three founders, three policies (Illustrative)

Illustrative. Generic companies with invented facts; no outcome is implied.

**Founder interest, no authority yet.** The co-founder and CTO of a 180-person field-services software company is keen after a portfolio dinner. Its policy sends any new external use of company data to an AI committee, and the CEO has not heard the idea. Status: hold. The CTO takes a one-page note to the CEO and the committee. You contact no one, and nothing is submitted.

**Authorized sponsorship with a narrow scope.** The CEO of a 420-person insurance claims administrator reads the map with counsel. Client contracts restrict claim files, so those stay out. The CEO authorizes a conversation limited to the company's own procedures and QA records from 2014 onward, and names the COO as internal sponsor. Status: introduce, with that scope written into the handoff note.

**A voluntary decline.** The founder and CEO of a 95-person HR software company rereads its public terms, which promise customers their content will not train AI, and says no. Status: closed. You thank them and do not raise it again.

## Keep the founder's enthusiasm and the company's decision apart

Write two lines, never one. The first records the founder's interest: who, when, what they said. The second records the company's decision: whether the CEO wants the conversation and whether any approval the policy requires exists. A founder-CEO can fill both, once the policy's approver has had their say; any other founder fills only the first. An introduction made without the decision maker's knowledge is, in SourceX's own words, usually not a fit. For the choice between going through a sponsor or straight to the CEO, see [introduce through a sponsor or directly to the CEO](/compare/introduce-through-sponsor-or-directly-to-ceo).

## The handoff note carries only what the CEO approved

When the CEO says yes, send SourceX a short note and nothing else.

- **Who and when.** The CEO's name and the date they said they want the conversation.
- **Scope at metadata level.** Record types, rough volumes, years and systems the CEO approved, such as "internal procedures and QA records since 2014, one system."
- **Clauses the company flagged.** Named, not attached: "client contracts restrict claim files" or "AI committee approval given on a stated date."
- **What stays out.** The policy document itself, client names and anything the CEO asked you not to pass on.

Sign the SourceX Channel Partner Agreement before you submit anything. Credit is decided from dated records once the company files a verified application, and it goes to the earliest valid introduction of that new company inside the 90-day attribution window, which is one more reason to keep the note.

## Fit facts the policy conversation cannot replace

A clean policy reading does not make a company a fit. The usual tests still apply. Its headcount must have peaked at 50 or more full-time employees, not counting contractors; its operating history must be documented; it must hold the rights to license its records; and its CEO must want the conversation. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English. Whether the records match a current program stays unknown until SourceX reviews them. For the basics, see [explaining data licensing to a founder](/guides/how-to-explain-company-data-licensing-to-a-founder); for ownership claims, [a founder says they own the data](/guides/a-founder-says-they-own-the-data-what-happens-next).

## Methods note and where the map can be wrong

**Method.** The nine clause types are an editorial composite built from three published sources checked on 2026-10-26: NIST's governance framing, the confirmations SourceX asks suppliers to make, and the program's screening profile. It is not a survey of company policies, and no founder interviews on this question have been published.

- **Bespoke policies.** A policy may combine rows, name unusual approvers or say nothing at all. The map tells the founder where to look, not what they will find.
- **Interpretation.** Only the company and its counsel can say what its policy and customer promises mean. The FTC post describes an enforcement position, not a court ruling.
- **Rights are decided in review.** A permissive policy cannot clear a client contract, and SourceX's rights check can still narrow or end a scope.
- **Programs change.** Criteria follow each program, and a company that misses one may fit another later.
- **Relationship cost.** Raising licensing with a founder already wary of AI spends goodwill; no data on founder reactions has been published.

## Next step

> **Refer a company. No cap on rewards.** Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.

When the company has read its own policy and the CEO wants the conversation with an agreed scope, create your partner link on the [registration page](/register), hold the company up against [who qualifies](/who-qualifies) and go through the [program terms](/terms) before you send the handoff note. Platform teams running this across a portfolio can start from the [venture platform overview](/for/venture-platform-teams) or browse the [guides](/guides).

Introducing a founder's company is referral work; deciding whether SourceX itself is worth investing in is a separate question, and the [SourceX investor overview](https://sourcex.si/investors) supplies investor materials to anyone who asks.

This article is for information only and is not an offer to sell or a solicitation of an offer to buy securities. Any offering by SourceX is made only to verified accredited investors through its offering documents.

## Sources

- https://refer.sourcex.si/terms
- https://refer.sourcex.si/who-qualifies
- https://sourcex.si/how-it-works
- https://www.ftc.gov/policy/advocacy-research/tech-at-ftc/2024/02/ai-other-companies-quietly-changing-your-terms-service-could-be-unfair-or-deceptive
- https://nvlpubs.nist.gov/nistpubs/ai/nist.ai.100-1.pdf
- https://delcode.delaware.gov/title8/title8.pdf
- https://sourcex.si/investors
