Singapore — PDPA (candidate data), AI decision-system guidance, and fair employment
What it asks of employers
- Get consent and tell candidates what you collect. On or before collecting a candidate's personal data, obtain their consent (or rely on a lawful exception) and inform them of the purposes for which it will be collected, used, and disclosed.
- Be transparent that you use AI to help make hiring decisions. Make your written data-protection policy say that you use an AI system to make or assist hiring decisions, the personal data it uses, and the safeguards you rely on — and make that policy available to candidates, not only on request.
- Correct inaccurate candidate information on request. When a candidate says information about them is wrong, correct it — in your applicant tracking system if the record is there, or by asking Amp to update information held in Amp, which Amp does through support on request.
- Hire on merit and don't let AI discriminate. Recruit and select on merit, and don't use AI in hiring in a way that discriminates against a candidate on a protected characteristic; review Amp's bias-test and audit evidence before and while you use the tool.
What Amp does
- Amp answers candidate requests to see their data. When a candidate asks for the personal data you hold, Amp gathers it so you can give the candidate access.
- Amp keeps candidate data only as long as it's needed. Amp follows your organization's retention policy and stops keeping candidate data once the purpose it was collected for — and any legal or business need — has ended.
- Amp bias-tests scoring and applies it consistently. Amp regularly tests candidate scoring for bias across demographic groups (by race and sex) and uses the same job-related scoring for every candidate — evidence you can use for merit-based, non-discriminatory hiring. Amp's bias test covers race and sex; other protected characteristics remain your own assessment.
- Evidence of how an automated decision was reached. Amp can produce a tamper-evident package documenting how the scoring works, which supports the transparency and accountability the AI guidance expects when personal data feeds an automated decision — and helps you explain one.