In this guide · 8 sections
- 01Is it legal to use AI in hiring in Singapore?
- 02Writing job descriptions with ChatGPT: what to check
- 03AI CV screening: where bias creeps in
- 04TAFEP's three rules for using AI in hiring
- 05PDPA: feeding CVs into an AI tool
- 06What the Workplace Fairness Act will add
- 07Do it yourself or get help?
- 08Frequently asked questions
Short answer
Yes, you can use AI for job ads and CV screening, but you stay responsible for a fair result. Check every AI-written ad against TAFEP's wording rules, feed tools only job-related criteria, tell candidates how AI is used, and let a person make the final decision.
Is it legal to use AI in hiring in Singapore?
Yes. Singapore has no law that bans or specifically licenses AI hiring tools. In a written parliamentary answer on 5 May 2026, MOM said the government is still studying how AI tools affect employment before deciding on an approach. In the same answer it was clear about today's position: employers must follow the Tripartite Guidelines on Fair Employment Practices (TGFEP) and the upcoming Workplace Fairness Act when they use AI tools.
In other words, "the software did it" is not a defence. If an AI screen filters out older applicants or a chatbot-written ad asks for "Chinese preferred", it is your ad and your process. Candidates can complain to TAFEP, and TAFEP works with the employer to fix it.
| Rule | Status (October 2026) | What it means for AI tools |
|---|---|---|
| Tripartite Guidelines on Fair Employment Practices | In force; all employers | Merit-based criteria, fair ads, consistent selection, whatever the tool |
| Workplace Fairness Act 2025 | Passed; not yet commenced on 19 Oct 2026 | Employers with 25+ staff; 11 protected characteristics |
| PDPA | In force | Notice and consent for CVs; vendor and overseas transfer checks |
| MyCareersFuture advertising (EP, S Pass) | In force | AI can draft the ad; the 14-day advertising rule still applies |
Writing job descriptions with ChatGPT: what to check
AI tools are good at turning a few notes into a tidy job description. They also reproduce the phrases found in millions of old job ads, some of which TAFEP specifically says not to use. Its page on writing job advertisements lists examples.
| Phrase AI drafts often include | Why it is a problem | Fair rewrite |
|---|---|---|
| "Young, dynamic working environment" | TAFEP lists "young/youthful working environment" as wording to avoid | "Fast-paced team with weekly deadlines" |
| "Age 25 to 30" or "recent graduate only" | Age as a criterion without a legal reason | "Diploma in accounting; training provided" |
| "Strong guys needed" | Gender wording TAFEP flags | "Must lift and carry 20kg cartons daily" |
| "Mandarin speaker" with no reason | TAFEP asks you to state the job-related reason for a language | "Mandarin needed to serve our China-based suppliers" |
| "Native English speaker" | Can read as a nationality or race preference | "Writes clear business English" |
A simple fix: tell the AI in your prompt to "use only skills, qualifications, experience and job conditions as requirements, and do not mention age, gender, race, religion, nationality, marital status or family responsibilities". Then read the result yourself anyway. Our guide to job ad rules for your careers page covers the full TAFEP list and MyCareersFuture rules.
AI CV screening: where bias creeps in
TAFEP's October 2025 article Fair Hiring First, AI Second names the risks directly: a tool that is not set up and checked properly can repeat bias from past hiring data, make it hard to see how a score was formed, and filter out qualified people with non-traditional profiles.
For a small business, the practical risks look like this:
- Copying your past hires. A tool that ranks CVs by similarity to "our best staff" will favour whoever you hired before, including any pattern you never intended.
- Proxies for protected traits. Graduation year reveals age. Some school names, clubs or addresses hint at race or religion. Remove them or tell the tool to ignore them.
- Career gaps. Automatic penalties for gaps tend to hit carers, often women, and people returning from illness.
- Hard knock-out filters. "Reject anyone without a degree" may screen out experienced people who could do the job. Use must-haves the job truly needs.
- Black-box scores. If you cannot say why a candidate scored 42 and another 78, you cannot defend the decision to TAFEP or to the candidate.
TAFEP's three rules for using AI in hiring
TAFEP's 2025 article gives employers three practical principles. Here they are as steps you can follow for each role.
Write the job requirements first, then configure the tool
Base the AI's criteria on a job analysis: the skills, qualifications and experience the role needs. Give the tool only that job-related data, not photos, age, nationality or marital status.
Keep people in control of every decision
A recruiter or hiring manager reviews the shortlist and anyone the tool scored low but who meets the must-haves. Final decisions rest with people, as TAFEP recommends.
Keep a record of each hiring stage
Note who was screened in or out and why. TAFEP says documentation supports accountability and gives clear evidence of how outcomes were reached.
Tell candidates how AI is used
Say in plain words what the tool does (screening, scheduling, shortlisting) and that people make the final decision. Give a contact for questions or appeals.
Spot-check the results
Every few months, compare who applied with who the tool shortlisted. If a group drops out sharply at the AI stage, look at the criteria.
PDPA: feeding CVs into an AI tool
CVs are personal data, so the PDPA applies the moment you paste one into a tool or upload it to an applicant tracking system. Two questions matter most.
Have you told candidates?
PDPC's advisory guidelines on AI recommendation and decision systems (issued 1 March 2024) say notifications should explain, in proportion to the risk, what the system does, what types of personal data it uses and how that data relates to the outcome. One line in your job ad or application form usually covers a small business: "We use software to help sort applications against the job requirements; a person reviews every shortlist."
What does the tool do with the CV?
PDPC's guidelines on personal data in generative AI (July 2026) treat the business that deploys an AI system as primarily responsible. A free consumer chatbot may keep inputs or use them for training. A business plan or an ATS with a proper data agreement usually does not. Check the vendor's data-use settings, where the data is stored, and how long it keeps it. Our ChatGPT at work policy guide covers the settings, and the recruitment agency PDPA guide covers ATS contracts and old CVs.
What the Workplace Fairness Act will add
Parliament passed the Workplace Fairness Act 2025 in January 2025. Singapore Statutes Online still listed it as not yet commenced on 19 October 2026. Once in force it applies to employers with 25 or more employees and covers 11 protected characteristics, including age, nationality, sex, marital status, race, religion and disability.
Nothing in the Act carves out decisions made with software. A screening rule that would be discriminatory if a person applied it is just as discriminatory when a tool applies it. Smaller employers below the threshold are still expected to follow TAFEP's guidelines, so the process above is the right default for everyone.
Do it yourself or get help?
Most small businesses can do this in-house: a fair prompt, a human review step, a line in the job ad, and a check of the AI tool's data settings.
Before your next AI-assisted hire
If you want help, our Security Awareness Training (S$390) covers safe use of AI tools with personal data for your staff, and the PDPA Compliance Pack (S$690) includes the notices, retention rules and vendor checks for candidate data. We do not give legal advice; for a specific dispute, contact TAFEP or a lawyer.
Frequently asked questions
No law in Singapore specifically requires it today, but TAFEP advises employers to explain in simple terms how AI tools are used and that people make the final decision. PDPA notification rules also mean you should tell applicants the purposes for which their data is used. A short line in the job ad or application form is enough for most small businesses.
Only if the account's data settings suit personal data. Free consumer accounts may keep conversations or use them to improve models. A business plan with training turned off and a clear data agreement is safer. Remove details the tool does not need, such as NRIC, photos and dates of birth, and keep a person reviewing the shortlist.
Not automatically. MyCareersFuture and job portals may reject some ads, but responsibility stays with the employer. TAFEP's guidelines say employers and their intermediaries should review all job advertisements before posting. Read every AI draft for age, gender, race, nationality and language wording before you publish it.
Once in force, most of the Act applies to employers with 25 or more employees; the threshold can be changed by the Minister. Smaller employers are still expected to follow the Tripartite Guidelines on Fair Employment Practices, so fair job ads and merit-based screening apply to every business, with or without AI.
Have a person review the application against the job requirements and record the outcome. Explain in plain terms how the tool was used and what criteria applied. If the criteria were not job-related, fix them for future roles. Candidates can also approach TAFEP, which works with employers to resolve complaints about unfair hiring practices.