Why the internal policy is not enough
A freelancer is not an employee, and your AI policy assumes one.
An internal AI policy assumes an employment relationship: HR onboarding, company devices, a badge that gets deactivated on the way out. None of that applies to a freelancer working on their own accounts, their own machine, on a per-assignment basis, often for other clients at the same time.
That gap matters most in three places: confidentiality (what stops your brand and client data from being pasted into a tool outside your control), ownership (who owns an AI-assisted deliverable once it is paid for), and disclosure (how you actually find out AI was used, when there is no manager watching over a shoulder). The addendum below closes that gap without replacing your internal policy; it sits alongside it.
The template
Attach this to the contractor's SOW
Fill in the brackets and attach it to the freelancer's statement of work or contract. It is an operations document, not a legal one; if enforceability matters to you, have your actual contract language reviewed by counsel.
FREELANCER AI ADDENDUM
Applies to: [contractor name]
Effective: [date] Reviewed: [date]
1. SCOPE
Applies to all paid work for [company] under the
agreement dated [ ].
2. APPROVED TOOLS
Use only these tools on our assignments:
- Approved: [tool], [tool]
New tools need [owner]'s approval before first use.
3. DISCLOSURE, EVERY INVOICE
Each invoice includes: tool(s) used, rough % of the
deliverable that was AI-assisted, and confirmation that
every claim was human-verified.
4. CONFIDENTIALITY
Client names, briefs, and any private data may not be
pasted into a tool outside the approved list. Approved
tools must have data retention/training disabled where
that setting exists.
5. OWNERSHIP
All delivered work, AI-assisted or not, is owned by
[company] once paid for under the underlying agreement.
No delivered content may be reused or resold to another
client without written consent.
6. PROHIBITED
- Fabricated sources, quotes, stats, or case studies
- Fully AI-generated sections submitted without disclosed
human review and rewrite
- Submitting the same AI output to more than one client
7. CONSEQUENCES
Undisclosed violation found after payment: right to
withhold future assignments and, per the underlying
contract, claw back the affected fee. Repeat violation:
contract termination.
8. ACKNOWLEDGMENT
[Contractor signature or written email confirmation]
[Date]How to use it
Tie disclosure to a checkpoint that already exists
Send it at onboarding, not mid-project. A freelancer who agrees to this before the first assignment has no argument later. Introducing it after work has started reads as a trust problem, even when it isn't one.
Tie disclosure to the invoice, not the draft. Freelancers already submit invoices to get paid, so riding that existing checkpoint costs nothing new to enforce. It also gives you real leverage: an invoice without the disclosure line is incomplete, not paid.
Get a written acknowledgment, even an email reply. The point is not building an enforcement case; it is that nobody can claim surprise about the disclosure requirement after the fact.
Keep the tool list current. An approved-tools list from six months ago is a list nobody follows. Review it whenever a freelancer asks to use something new, and say yes or no in writing.
When this stops being enough
One addendum works for one freelancer at a time.
This template is enough for a handful of contractors on informal terms. It strains once you have five or more freelancers whose disclosures need tracking over time, once the addendum needs to live inside a formal SOW template instead of a standalone attachment, and once freelancer and internal staff rules need to visibly match so nobody can point to a double standard.
At that point the addendum needs the rest of an AI operating layer around it: the internal policy it pairs with, an editor QA scorecard that catches undisclosed use, and a tracker that shows disclosure compliance across the whole roster.
The complete version
AI Ops Playbook for Content Teams
This addendum is one component of the playbook, built and maintained as one kit: the internal AI use policy, a copy/paste AI block for briefs, this Freelancer AI Addendum, an editor QA scorecard with automatic-stop rules, a red flag library, a source/claim worksheet, a client-facing AI blurb, and a lightweight AI Use Tracker. It includes a 60-minute pilot path so the rollout starts on your next three assignments, not next quarter.
It costs $37, comes in Word, Markdown, and CSV formats, includes free updates forever, and carries a 14-day fit guarantee: run the pilot for two weeks, and if it does not fit your workflow, email me for a refund.
Related resources
The internal counterpart
This addendum pairs with the AI Use Policy Template for Content Teams, the internal version for staff. Send both together so freelancers and employees work under matching rules, and use the Content QA Scorecard Template to judge AI-assisted drafts against the same standard as everything else.
FAQ
Common questions
How is this different from the internal AI use policy?
The internal policy covers staff workflow: approved tools, review steps, escalation. This addendum covers what only applies to an outside contractor: confidentiality once your data leaves your systems, ownership of what you are paying for, and disclosure tied to invoicing rather than an internal review step. Use both, not one instead of the other.
Do freelancers need to sign this separately from their contract?
A written acknowledgment is enough; a full signature is not required unless your standard contracting process already requires signed amendments. The goal is a clear, dated record that the freelancer saw and agreed to it before work started.
What if a freelancer refuses to disclose AI use?
Treat a missing disclosure line as an incomplete invoice, not a paid one, per section 3. Refusal after that is a fit problem worth addressing directly before the next assignment, not something to let slide because the work itself was good.
Does this apply to fully AI-generated first drafts?
Yes, if your internal policy allows AI-generated first drafts at all. Section 6 requires disclosed human review and rewrite for any AI-generated section; an undisclosed AI-generated draft submitted as original work is a prohibited use regardless of how it was framed to you.