An AI-resistant hiring process combines identity verification, structured assessment, independent background checks and controlled onboarding so that employers do not rely on CVs, video interviews or documents alone. Singapore organisations should verify the person, test the claimed skills, confirm material credentials and delay sensitive access until key checks are complete.
AI has changed the scale of candidate misrepresentation
Generative AI can help genuine candidates improve grammar, prepare for interviews and communicate their experience more clearly. It can also make fabricated applications faster and more convincing.
A person can produce a polished CV tailored to a job description, generate plausible project descriptions or rehearse answers with an AI assistant. More sophisticated fraud may involve altered documents, synthetic profile images, voice manipulation or another person participating in an assessment.
The correct response is not to reject every AI-assisted application. Employers need controls that focus on evidence and consistency. The question should be whether the candidate is genuinely the person being assessed and whether the material claims influencing the decision are true.
1. Define the risks before choosing checks
Start with the role. Consider access to money, customer data, source code, critical infrastructure, vulnerable people and regulated activities. A role-risk assessment helps the employer choose proportionate controls rather than treating every applicant as equally risky.
Define which credentials are mandatory, which experience claims are material and which checks must be completed before the person starts. This creates a defensible process and reduces the temptation to add intrusive screening after a suspicious interview.
2. Verify identity early and consistently
Ask candidates to use the same legal identity throughout the process and record former or alternative names appropriately. Identity verification can include document authenticity checks, liveness measures and comparison with the person attending the interview.
Remote candidates may require an additional live identity step. Recruiters should watch for unexplained changes in appearance, audio, location, contact details or communication style, while avoiding assumptions based on accent, disability or technical quality alone.
Reconfirm identity before issuing equipment or privileged access, particularly where the hiring process has been entirely remote.
3. Use structured, work-relevant assessment
AI can generate credible interview answers, so unstructured conversation is a weak control. Use consistent questions linked to the role and ask candidates to explain decisions, trade-offs and examples in depth.
Work samples, technical exercises and scenario discussions can reveal how a person approaches real problems. Assessors should ask follow-up questions about the candidate’s reasoning and adapt the scenario rather than relying on a static test that can be copied into an AI tool.
Document scoring criteria in advance. This improves fairness and makes it easier to distinguish evidence of ability from presentation quality.
4. Verify the claims that matter
Background screening should confirm the information that materially affected the decision. This may include employment dates, job titles, qualifications, professional licences and directorships.
For candidates requiring an Employment Pass, determine whether qualifications will be used for COMPASS points and obtain acceptable MOM verification proof. A professional verification should use authoritative sources rather than judging a certificate only by appearance.
Discrepancies should trigger review, not an automatic conclusion. Give the candidate an opportunity to explain and apply a consistent decision framework based on relevance, seriousness and evidence.
5. Join screening to onboarding
Fraud controls weaken when recruitment ends and onboarding begins. HR, IT and the hiring manager should share a clear status for required checks.
Use conditional offers and define which controls must be complete before system access, financial authority or customer contact is granted. Where an urgent start is commercially necessary, apply temporary restrictions and record who approved them.
Confirm that the person who joins is the person who completed the process. A brief manager-led identity and experience discussion on the first day can help identify unexplained inconsistencies.
6. Train recruiters without creating a culture of suspicion
Recruiters need practical guidance on synthetic identities, altered documents and AI-assisted interviews. They should know how to escalate concerns and preserve evidence.
Training should also emphasise fairness. Nervousness, poor connectivity or an unfamiliar communication style are not proof of fraud. Controls should be objective, repeatable and respectful.
Explain the process to candidates. Clear communication about identity and background checks can deter misrepresentation while reassuring genuine applicants that everyone is assessed consistently.
Build resilience through layered evidence
No single interview question, identity tool or background check can eliminate hiring fraud. Resilience comes from several controls supporting one another: verified identity, structured assessment, independently confirmed claims and controlled access.
RMI can support Singapore employers with managed identity, employment, education and MOM verification. By designing the process around evidence rather than appearances, organisations can use AI productively while reducing the risk that technology is used to conceal who a candidate is or what they have actually done.