# An AI use policy for HR: the policy, the tool register and the rollout

AI USE POLICY KIT, read 10 Nov 2026

**The short answer.** An AI use policy tells employees which AI tools they may use, what data may go into them, and when a person must check the output. The kit has the policy, a tool register, three data tiers, a rule for employment decisions and a 30-day rollout.

## The policy

An AI use policy lists approved tools, banned data, human review and disclosure. The key US fact is that no single federal AI hiring law exists, while New York City requires a bias audit and notices for automated hiring tools. In the UK, solely automated significant decisions need safeguards. Companies choose which tools are approved.

**POL-38 AI use policy**

### Purpose

This policy sets how [Company] uses artificial intelligence tools so that work is faster without exposing people, customers or confidential information. Anyone may ask [contact] if they are unsure whether a use is allowed. We would rather answer a question than repair a mistake.

### Who this covers

All employees, contractors and interns who use AI tools for [Company] work, on company or personal devices.

### Approved tools

You may use only the tools on the approved list kept by [owner, for example IT lead]. To request a new tool, send the name, the use and the data involved to [contact]. Free accounts on public tools are not approved for work data. The list shows, for each tool, the data classes it may hold and who may use it. [Owner] reviews the list every [number] months.

### Data you must not enter

Do not enter personal data about employees, candidates or customers, health information, passwords, source code marked confidential, or anything covered by a customer contract, unless the tool is approved for that data class.

### Human review

A person is responsible for every output used in work. Check facts, sources and tone before sharing. Never use AI output alone to hire, reject, discipline, promote, pay or dismiss anyone. A trained reviewer must make and record those decisions.

### HR and hiring uses

Before any tool screens, ranks or scores candidates or employees, People Ops must approve it, confirm what bias testing exists, and give candidates and employees the notices the law requires. Candidates may ask for a person to review their application.

### Disclosure

Tell [customers, colleagues, candidates] when content or a decision was produced mainly by AI if they would reasonably want to know. Do not present AI-generated work as your own where a customer, regulator or school expects original work.

### Reporting and consequences

Report suspected misuse or a leak within [number] hours to [contact]. Breaking this policy can lead to disciplinary action up to dismissal.


## What the law requires

| Rule | Jurisdiction | Law and source |
| --- | --- | --- |
| No federal statute regulates employer use of AI in hiring; the general ban on practices with an unjustified disparate impact on protected groups applies when a tool is used to select people. | US federal | [Title VII of the Civil Rights Act, 42 U.S.C. 2000e-2(k) (disparate impact)](https://www.law.cornell.edu/uscode/text/42/2000e-2) |
| An employer may not use an automated employment decision tool for hiring or promotion in New York City unless it had a bias audit within one year, a summary of the audit is public, and candidates and employees received the required notices. | New York City | [NYC Local Law 144 of 2021 (Automated Employment Decision Tools), as described by the Department of Consumer and Worker Protection](https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page) |
| A significant decision about a person based solely on automated processing, with no meaningful human involvement, must have safeguards: information about the decision, a way to make representations, human intervention on request, and a way to contest it. Stricter limits apply where special category data is used. | UK | [UK GDPR Articles 22A to 22D, substituted by Data (Use and Access) Act 2025, section 80](https://www.legislation.gov.uk/ukpga/2025/18/section/80) |
| EU row, for companies hiring in the EU: AI systems used for recruitment, selection, decisions on work terms, promotion, termination, task allocation or monitoring and evaluation of workers are listed as high-risk, and an employer using one must inform workers' representatives and the affected workers before use. | European Union | [Regulation (EU) 2024/1689 (AI Act), Annex III point 4 with Article 6(2), and Article 26(7)](https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689) |

The US rows come first, then the UK rows. Each row was read on 10 Nov 2026 on the page linked.

State and city laws add to this federal floor; check the state labor department's page for each state where you employ people.

## The tool register

Keep one list of the AI tools the company has approved. A tool that is not on the list is not approved for work data. Review the list every quarter.

| Column | What goes in it |
| --- | --- |
| Tool and vendor | The product name and the company that supplies it |
| Approved use | The tasks it may be used for, for example drafting job ads or summarising meeting notes |
| Data tier allowed | The highest data tier that may be entered, from the tiers below |
| Owner | The person who answers for the tool, usually the head of the function that asked for it |
| Approved on and review date | When the tool was approved and when it is next reviewed |
| Human check | Who checks the output before it is used, and for which uses |

A register is a table. A spreadsheet is enough at 50 people.

## Three data tiers

| Tier | Examples | AI tools |
| --- | --- | --- |
| Public | Published job ads, the careers page, public policies | Any approved tool |
| Internal | Org charts, internal memos, draft policies | Approved tools with a contract that keeps the data out of training |
| Personal or confidential | Employee records, pay, health and disability, performance reviews, candidate files | Only a tool named in the register for that use |

## Employment decisions need a person

Set a rule that no hiring, pay, promotion, discipline or dismissal decision is made by an AI tool alone. A named person reviews the output and records the decision. The EU AI Act classes AI used for these decisions as high risk; the next page gives the article.

## A 30-day rollout

1. Days 1 to 5: list the AI tools already in use, by asking each team.
2. Days 6 to 10: approve or retire each tool and fill the register.
3. Days 11 to 15: fix the data tiers and get the policy reviewed by a lawyer.
4. Days 16 to 20: publish the policy and send it for acknowledgement in the HR system.
5. Days 21 to 30: run a short briefing for managers and check who has not acknowledged.

## Standards and law

- [The EU AI Act and employment uses](/ai-use-policy/eu-ai-act/): Annex III point 4 and the deployer duties
- [ISO/IEC 42001 in one page](/ai-use-policy/iso-42001/): The AI management system standard
- [The AI use policy page](/policies/ai-use-policy/): Template, legal floor and the platforms that can enforce it

## Sources

- US federal: Title VII of the Civil Rights Act, 42 U.S.C. 2000e-2(k) (disparate impact): https://www.law.cornell.edu/uscode/text/42/2000e-2 (read 10 Nov 2026)
- New York City: NYC Local Law 144 of 2021 (Automated Employment Decision Tools), as described by the Department of Consumer and Worker Protection: https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page (read 10 Nov 2026)
- UK: UK GDPR Articles 22A to 22D, substituted by Data (Use and Access) Act 2025, section 80: https://www.legislation.gov.uk/ukpga/2025/18/section/80 (read 10 Nov 2026)
- European Union: Regulation (EU) 2024/1689 (AI Act), Annex III point 4 with Article 6(2), and Article 26(7): https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689 (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://freeemployeehandbooktemplate.com/ai-use-policy/
